2. 3 // Multichannel attribution, contribution & optimisation
June 2016
Acknowledgements
The EBG would especially like to thank members and partners of the “ Multichannel attribution,
contribution and optimisation ” Task Force: Performics. A big thank you to the Performics / Publicis Media
teams who helped us to set up and grow this project: Pierre-Emmanuel Cros, Laure Debos, Aude Guérin, Arnaud
Lauga, Nicolas Schmitz and Guillaume Valicon, as well as Florence Delplanque-Boulenger who compiled
the information and formatted this publication.
With support from iStock by Getty Images
Multichannelattribution,
contribution&optimisation
What is a Task Force?
It is a discussion process limited to around 20 decision makers
for major accounts, focusing on four objectives:
1. Discussion among peers
2. Making headway on a strategic topic
3. Identifying reliable partners
4. Sharing ideas with the whole EBG community
Between October 2015 and May 2016, the “ Multichannel attribution,
contribution and optimisation ” Task Force, organised by the EBG
in partnership with Performics, held a discussion on current attribution
models and their development in the near- and medium-term. How can one
measure real return on each of the investment levers for both online and
offline channels? What impact do levers have on one another? To what extent
do they feed on one another? What is the right amount to invest, overall
and for each channel? How can I measure the incremental contribution
of media, whether in terms of traffic, sales or brand image?
The answers are in this book.
THE ELECTRONIC BUSINESS GROUP - www.ebg.net
The EBG is the largest French professional community for company decision-
makers in industry, services and media, who all share an interest in innovation.
The EBG’s activities:
The purpose of the EBG is to maintain a knowledge and know-how network
by organising more than 150 round tables every year, giving senior executives
the chance to share their experience.
There are more than ten active communities within the EBG: CEOs, Purchasing
Directors, IT Directors, Finance Directors, HR Directors, Marketing Directors,
Web Marketing/e-Commerce Directors, Media and Mobility Directors,
and Communication Directors.
Each community meets at least once a month for discussions and updates on
the newest practices in their field. The sessions are filmed and streamed live.
For more information on the EBG or its partners:
Benjamin Glaesener
Managing Director
+33 1 45 23 05 89
benjamin.glaesener@ebg.net
Fanny Raimbault
Content Manager
+33 1 73 03 01 98
fanny.raimbault@ebg.net
ACCORHOTELS
Air France
Allianz
APEC
April Group
AramisAuto.com
Auchan Direct
BNP Paribas Cardif
Celio France
Compagnie des Alpes
CRM Services filiale SNCF
DS AUTOMOBILES
(PSA PEUGEOT CITROEN)
Euromaster
Fortuneo
FUTUROSCOPE
LA BANQUE POSTALE
LA FRANCAISE DES JEUX
LA POSTE
LUDENDO
LVMH / Moët-Hennessy -
Louis Vuitton
MALAKOFF MEDERIC
Meetic - MEETIC MOBILE
MeilleurTaux.com
MICHELIN
ORANGE / FRANCE TELECOM
ORANGE BUSINESS SERVICES
PMU
RENAULT
SFR
SONY MOBILE
COMMUNICATIONS FRANCE
Ubisoft
VANITY FAIR BRANDS
Lingerie SA
members of the task force
3. [ This guinea pig is the EBG’s
mascot. He will be your guide
through this publication
and will provide details, figures
and additional information.
And if you stop by to see us
at 10 rue Mercœur (Paris,
11th
arrondissement), you can
meet him in person. ]
5 // Multichannel attribution, contribution & optimisation4 // Multichannel attribution, contribution & optimisation
Beyond “ last click ”
(page 6)
Measuring performance
The objective: a sale, a lead
and more
Six big attribution models
Case studies
Sony Mobile: “ Service
providers versus in-house
solutions ”
(p. 12)
L’Oréal: a complicated
network and concrete
answers
(p. 14)
Spotlight
Social networks are
jealously guarding
their turf
(p. 16)
table of contents
#1 #2 #3
[ Chapter 1: understanding attributi
on-contribution ]
[ Chapter 2: reconciling two differen
tworlds ]
[ Chapter 3: in-depth look at manag
ementbyincrements ]
Excellingatmultichannel
(page 30)
Cross-device and the
dream of the unique ID
Incremental management
Case studies
Compagnie des Alpes: bi-
annual attribution studies
(p. 36)
Greenweez.com:
“ Identifying value-creating
channels ” (p. 38)
Motoblouz.com: “ Avoiding
technological one-
upmanship ” (p. 40)
Spotlight
First steps towards
incremental management
(p. 42)
Bringingintheofflineworld
(page 18)
Different calendars
and units of measurement
Focus on econometrics
The tenuous link between
Online and Offline
Case studies
Meilleurtaux.com:
“ Measuring the impact
of our TV spots ”
(p. 24)
Euromaster: “ The digital
world working for the
physical network ”
(p. 26)
Spotlight
DMP and CRM: towards
a more pragmatic approach
(p. 28)
4. 7 // Chapter 1 // beyond “ last click ”6 // Chapter 1 // beyond “ last click ”
occurs through paid channels,
while the other half leverage
the attribution model that cre-
dits the last click, regardless
of channel. Moreover, Google
Analytics and other site-centric
tools and ad servers are confi-
gured by default to incorporate
the last click”.
Members of the EBG 2016 Task
Force working on “ multichan-
nel attribution, contribution
and optimisation ” were aware
of this, stating in the very first
session that “ analysis by ‘last
click’ undervalues branding
performance at the expense of
the performance of the SEA and
proprietary levers ”.
What impact do levers have on
one another? To what extent do
they feed on each other? How
can you measure the incre-
mental contribution of a media,
whether in terms of traffic,
sales or brand image? At what
cost is this contribution made?
Getting more sophisticated
These questions are essential
and lead to others about the
offline world and about the uni-
queness of clients: what impact
do digital campaigns have on
conversions in the physical
world? Is it possible to include
offline marketing campaigns in
attribution models? How can one
identify a single client through
various devices (telephone, com-
puter, tablet) and the different
channels they leverage?
The stakes are high. “ These
days, we consider the contri-
bution of each ‘device’ and
each channel to overall traffic ”,
explains Yoann Aldon, Res-
ponsible for Digital Strategy
and Client Awareness at the
Française des Jeux. “ We have
around 33,000 sales outlets
in France, and we are not ex-
clusive at these locations. We
therefore have a major problem
regarding Online vs Offline in
terms of intermediation and
recovering information ”.
For Meetic, the European leader
in online dating (Meetic, Friends-
cout, Match, Lexa, Neu, etc.),
online acquisition “ is naturally
at the heart of our work, and
comes with specific cross-de-
vice attribution problems. As a
pure player, we are very aware
of Online, but we also have a
strong Offline presence and are
currently working on an in-
house attribution model that will
be more accurate for TV spots in
particular ”, says Pierrine Grif-
fiths, Mobile Services Manager.
Attribution is also a priority for
a certain major players in shoe
sales: “ We started up in digital
relatively late, and I am current-
ly in the process of establishing
centralised management. Ha-
ving more sophisticated attribu-
tion and contribution models is
a priority ”.
“ Why bother with attribution?
Generally speaking, it’s to mea-
sure the profitability of your
marketing actions ”, says the
Performics team. “ Different
attribution models exist these
days, with increasingly scientific
approaches: mathematics and
even statistics ”.
Before going into detail on
these models, it is important
to define certain terms: what is
meant by “ attribution ”, “ contri-
bution ” and “ deduplication ”?
The Performics team: “ Attribu-
tion is attributing a conversion
to a particular lever, accor-
ding to a predefined rule set.
Attribution is challenged by
contribution. Contribution is
about identifying which stages
were decisive and what role
each lever performed. As for
deduplication, this ensures that
each conversion is only counted
once. This is a prerequisite for
both attribution and contribu-
tion ” (continued on the next
page).
G
oogle “ last click is dead ”
and you will find that the
death of the last click attri-
bution model was announced at
least eight years ago, in 2008.
Nonetheless, the latest sta-
tistics from IFOP, the French
Institute of Public Opinion find
that the vast majority of French
advertisers still measure their
performance by the last click:
79 %, according to a study car-
ried out in late 2014.
The reason is obvious: it is the
easiest and most accessible
model. However, advertisers
are not naive. The last click is
an illusion.
“ There are some cases - maybe
one study in five - where attri-
bution by last click is accurate ”,
according to Arnaud Lauga,
Data, Technology & Innovation
Director at Performics/Publi-
cis Media France. “ Four out
of five advertisers base their
models on the ‘last click’: half
of these ‘last click’ advertisers
use an attribution model that
gives credit to the last click that
zoom
14 marketing actions for a conversion
According to an analysis carried out by Performics in 2016,
it takes 14 marketing actions to create a digital conversion
(across all sectors). This is an average: more than half of digital
conversions are accomplished through a single action, with a
central tendency that has been relatively stable over the past
few years: between 40 % and 60 %. In these cases, attribution and
contribution become the same.
“ Analysis by last click undervalues branding
performance at the expense of the performance
of the S.E.A. and proprietary levers. ”
beyond “ last click ”
Towards a customised model.
chapter 1
5. special report
9 // Chapter 1 // beyond “ last click ”8 // Chapter 1 // beyond “ last click ”
W
hile “ performance ” is
quintessential, it does not
represent the same thing
to everyone. Even the members
of the EBG Task Force have
different views on what
constitutes “ performance ”.
For Apec, performance refers to
appointments, while for Meetic,
it refers to recruiting free or
subscription-paying members.
The primary objective for
Auchan, a retailer that sells
staple products, is to improve
its understanding of customers:
its performance indicators deal
with recurring purchases and
average shopping baskets.
The objective: a sale, a lead
and more
Celio is making the transition
from an online acquisition
strategy to a holistic strategy,
since its 500 stores “ clearly
remain its primary business
contributors ”. An e-reservation
system meets these
requirements by allowing the
company to monitor in-store
sales as well as consumer
behaviour after its TV spots are
aired.
The most common metaphor
to explain the interest in
contribution comes from sport:
“ Of the marketing actions
you take, which levers are
completely autonomous? Which
are the initiators, the passers
and the strikers? ”
A sports metaphor...
or perhaps a criminal one?
The Performics team illustrates
its methods differently:
“ Imagine that a conversion
is a murder and you need to
identify those responsible. The
person who pulled the trigger
is guilty of course, but what
about the person who provided
the weapon or who ordered the
hit? Aren’t they also to blame?
The goal is to understand to
measuring
performanceA problematic idea! Performance does not mean the same thing to all
advertisers.
what extent each person was
responsible for the murder, and
to what degree the crime might
have taken place without the
gunman. If it was a paid hitman,
would the murderer have been
equally successful? ”
“ The goal of this investigation
is to inform the judge, who
interprets various laws to make
the right decision, attributing
blame and passing sentences
on each person involved in
the crime in proportion to
their level of responsibility.
Finally, the laws the judge
rules on are fairly subjective,
or at least bound by a certain
culture or era. That is why
these rules must be determined
carefully, so that attribution
for a sale through a lever or an
advertisement is proportional
to the real responsibility of this
action for the final conversion ”.
The key: deduplication
“ In the context of this crime
metaphor, deduplication
is the equivalent of having
five different investigators
investigating five different
suspects, says Guillaume
Valicon (Performics). In the
evening, they go back to the
police station and don’t speak
to each other. Because they
don’t speak to each other, we
might be led to think there
are five different murders
instead of one. Deduplication
is about communicating and
understanding that there is only
one conversion being discussed.
When we deduplicate, a single
conversion is attributed to
the various levers, along with
a proportion. Of course, to
get to this point, one needs
a customised measurement
ecosystem ”.
Six big attribution models
There are six classical
attribution models:
The “ last click ”
model gives all
the credit to the last point of
contact before the conversion.
This is the basic attribution
model and the default setting
for tools on the market. It is
an acceptable solution only for
short purchase funnels or for
impulse purchases.
The “ first
click ” model, in
contrast to the previous model,
gives all the credit to the first
point of contact. This favours
upstream marketing channels
and discounts the brand’s
marketing actions.
The
“ linear ” model gives each point
of contact the same value with
regard to the conversion. This
is used in cases of extensive
searches for high-value
products.
The model
based on the
importance of the first and last
actions can be adapted when
there are multiple searches and
points of contact, in order to
give more weight to the points
of contact that begin and end
the process.
It is very rarely used in practice.
The “ time
decay ”
model
gives most credit to the most
recent points of contact.
This is used for impulse
purchases for which the most
recent points of contact are
likely to have had the most
impact.
Finally,
customised
models, which are the most
accurate: “ These are the most
interesting models, where the
type of action is accounted
for, not just its position in
the purchasing funnel. These
models are more algorithmic,
more adaptable and therefore
more accurate than the
previous five. I don’t mean to
say that they recreate reality,
just that they are the most
accurate ”, says Arnaud Lauga
(Performics).
Apart from the models, the
tools currently available on the
market can be divided into three
main categories:
- “ Site-centric ” tools, such
as Eulerian, Google Analytics,
TagCommander, AT Internet
and Adobe Analytics. “ The
advantage of these tools is
that all the levers are available
natively after the click. These
tools can be used to determine
the remuneration of display
performance partners ”, says
Arnaud Lauga. “However, post-
view is only possible if bridges
or double tracking are in place.
The attribution models
« An sed audire aperiri, et quod amet consectetuer sea, no definiebas
adversarium sea. Sint percipit no nec, id vis viris aeterno euismod. Atomorum
pericula no qui, esse iuvaret usu ad. Ius ferri prompta praesent ut. »
[ Nullam semper ei eam ]
“The key to our approach is to compare conversion and non-conversion
pathways. This allows us to quantify the additional conversion probabilities
of each type of action ”.
[Arnaud Lauga, Data, Technology & Innovation Director at Performics/Publicis Media]
lll
[ Meilleurtaux.com (left) has come up with a tool to measure the impact of TV campaigns on site traffic: see page 24. Allianz
(right) is trying to optimise its Web to Store strategy.]
6. are (too) simple, and the
raw information is not always
exportable. Moreover, there are
substantial licensing costs for
some of these tools ”.
- “ Ad servers ” such as
DoubleClick Campaign
Manager, Sizmek and
Weborama. “ They integrate
post-impression and post-
view, providing exportable raw
files and advanced attribution
models ”, continues Arnaud
Lauga. ” However, they require
that the channels are double
tracked, which comes at a
cost. They do not allow onsite
unless special tracking is put
into place. Moreover, they
do not resolve the problem
of tracing non-paying levers
(except for ad hoc technological
developments) ”.
- Finally, “ attribution
specialists ” such as Mazeberry,
Visual IQ, Adometry, Abakus,
C3 Metrics and Converteo offer
advanced statistical models and
information visualisation. The
main downside: their cost.
“ The key to the Performics
approach is the comparison
between conversion pathways
and non-conversion pathways ”,
explains Arnaud Lauga.
“ This comparison allows
us to quantify the additional
conversion probabilities that
each action type provides.
From a technical standpoint,
we draw on a statistical
model: logistic regression.
Some models retrieve 95 %
of conversions and non-
conversions compared to the
control sample. We therefore
get a better understanding
of the ecosystem in terms
of conversion pathways,
non-conversion, interactions
and lever complementarity,
post-view windows and more.
This lets us measure the
contributions of branding
campaigns and acquisition,
adjust the associated costs and
challenge the different display
partners, removing some from
media plans and introducing
others we didn’t think were
profitable ”.
Limitations of the exercise
Of course, attribution/
contribution has its limitations.
It does not provide a perfect
understanding of attribution,
just more accuracy than what
is provided by default by other
tools. Another key question:
how to integrate Offline in an
attribution model largely based
on the digital world. Consumers
(and especially Internet users)
generate a large volume of
data. Reconstructing their paths
on the Internet - and in the real
world - is a major goal.
“ We at Boursorama are lucky
to be a pure player, and a small
brand that is growing quickly ”,
explained Nicolas Montetagaud,
Marketing Director, during an
EBG conference in 2014.
Imagine that a client hears
about Boursorama on
television, then discovers a
special offer while browsing
on the Internet, clicks and
then continues to surf. A
little while later, they search
“ inexpensive bank ” on Google
before arriving on the Web
page of a Boursorama affiliate.
This confirms the client’s
preconception of the bank and
clinches the sale: this affiliate
may now become the trigger,
even if the decision was not
necessarily made at this stage.
Certain solutions are being
developed to provide a pathway
between Online and Offline,
such as Atlas, DoubleClick,
and methods like CRM
onboarding and Call Tracking
(more information on page
22). These solutions all suffer
from the same shortcoming:
the providers have a conflict of
interest (see page 19).
Finally, imagine a world without
advertising: what proportion
of sales would occur anyhow?
The answer to this important
question, asked time and
again by members of the Task
Force, can be found through
“ incremental management ”
(see page 31).
11 // Chapter 1 // beyond “ last click ”10 // Chapter 1 // beyond “ last click ”
lll
[Opposite: display from the Meetic campaign “ Love Your Imperfections ”, launched in
2015 on TV, radio and digitally, and repeated in 2016. This campaign played on both
Online and Offline, including highlighting the dinners and workshops organised by the
pure-player in the physical world.]
zoom
Should one be opposed to the integration of branding
and performance?
Is it preferable - and possible - to integrate branding in
attribution/contribution methods?
At the Futuroscope theme park, the goal was to “ analyse the
contribution of each point of contact to the purchase ”. Florence
Lemesle, Head of Advertising and Media Partnerships, explains:
“ The difficulty in the leisure and travel sector is that purchase
recurrence is very low and the client funnel is very long.
Attribution windows can therefore extend to 90 days. We place
considerable importance on branding, which has an impact on
our objectives. Therefore, the visibility rate is a very important
performance indicator for us and is in no way excluded from the
attribution model... ”
Arnaud Lauga (Performics) emphasises that: “ Attribution is
identifying the advertising stimuli that accounts for a behaviour,
whether a purchase conversion or a psychological change
(awareness, image perception). We often speak of attribution for
a lead or sales action, but it can refer to a change in awareness of
your brand ”.
special report
“ Finally, imagine
a world without
advertising:
what proportion
of sales would
occur anyhow? ”
7. case study
13 // Chapter 1 // beyond “ last click ”12 // Chapter 1 // beyond “ last click ”
keypoints
information allows the company
to refine its digital strategy:
“ The goal is to bring the
customer to our own channels,
and in particular to our website
and the Xperia Lounge, two
networks where we can hold
a one-on-one conversation
without having to pay for it. After
all, on Facebook and Twitter,
the more you communicate, the
more you pay. We have about
50 million fans across all social
networks, and an unsponsored
post only allows us to reach
1 % of this fan base due to new
social network algorithms ”.
“ Making the most of three
different paths ”
In terms of tools and solutions,
David Chapon believes he
is at a crossroads between
three major paths: the “ major
players ”, with whom he has
been working for a few years,
smaller service providers, and
“ in-house ” solutions.
“ The major players offer an
impressive suite of programs,
but it costs a fortune and you
sometimes feel like you’re
playing the role of guinea pig.
The smaller providers are
certainly very interesting, but
working with them requires
hand-holding by an SS2I and
system integrators, since
we are forced to resort to an
amalgam of solutions: one
provider for social networks,
another for mobile Web
pages, etc. Connecting these
solutions together can be very
complicated ”.
The third way - an in-house
solution - is tempting: “ This
relates mostly to groups like
ours, with a strong desire to
innovate, the necessary human
resources internally (as a
smartphone manufacturer, we
have the required personnel)
and, especially in our case, a
major emphasis on protecting
private information. We are
particularly careful when it
comes to sharing our data with
service providers, whatever
their size ”.
“ The main risk for those
who choose to develop their
own solutions is long-term
maintenance. You can’t skimp
on it. We are currently trying
to leverage the best of three
worlds. In practice, there is no
tool that can do everything. It’s
a real problem ”.
Another difficulty mentioned
by David Chapon: the lack of
marketing support, especially
from social networks, which
“ tend to focus on advertising,
but should help brands by
offering marketing tools that
help develop one-on-one
conversations outside the
network: I think this would be
very smart in the long term ”.
As for Offline, David Chapon
is working on geo-tracking,
among other things, in order
to push offers to users who
pass by a store in order to sell
upscale products to existing
clients.
As a smartphone
manufacturer, Sony
Mobile has special access
to the screens of their
smartphones, “ which
consumers look at 50 to 70
times per day ”.
The group uses attribution/
contribution to build loyalty,
employing multichannel
while working on an in-
house solution in order
to protect its information
and to make up for the
inadequacies of solutions
on the market.
“M
any brands spend a
fortune trying to reach
smartphone screens,
which we have the opportunity
to access directly ”, says
David Chapon, Digital, CRM
and E-commerce Director for
Sony Mobile. “ We work with
substantial data, in terms
of both quantity and quality.
With regards to attribution/
contribution, we decided to
work on multichannel and
loyalty-building. We are looking
to optimise client pathways
on sites, mobile apps, social
networks and even television on
certain apps ”.
By accessing the data
anonymously, David Chapon
can find the consumer’s point
of entry into the Sony Mobile
ecosystem and reconstruct
their pathway. “ For example,
we now know that when a client
or prospective client uses at
least two pathways, we have a
50 % higher chance of retaining
or converting them. Therefore,
an existing client who receives
our emails but never goes to
our website will be harder
to retain and to sell upscale
products to ”.
This valuable type of
In order to attract customers to sonymobile.com and the Xperia Lounge,
Sony Mobile is sending an increasing number of push notifications throughout the client
life cycle and has adopted “ a media approach that includes regular communications
on news and contests... ” The threshold is a maximum of two push notifications
per week, but this may vary if the user reacts positively or indicates they are interested
in a particular field (such as sport, for example): in this case, notifications are
personalised and prominent.
As for the site, it is beginning to attain “ a high level of personalisation: it changes
when someone logs in, or if it recognises a cookie ”.
[David Chapon is Digital, CRM and
E-commerce Director for Sony Mobile]
Serviceprovidersversus
in-housesolutions
Sony Mobile
zoom
The figures
• Xperia Lounge has more than 50 million installations,
13 million active users (more than 2 visits per month) and
a 4/5 score on Google Play (also available for iPhone on
xperialounge.com)
• Sonymobile.com has more than 70 million visitors per month
and 52 localised versions of its website.
[75 % of website traffic comes from a mobile device.]
8. 15 // Chapter 1 // beyond “ last click ”14 // Chapter 1 // beyond “ last click ”
zoom
our branding strategies.”
With regards to attribution,
Vincent Labasse’s objective is
to attribute simple consolidated
and timestamped performance
indicators to each point of
contact with customers. “In
order to reconcile our offline
and online data and reduce
confusion, we have had to work
on several projects at once”.
Consolidating the data
First lever for the brand:
managing data better. With
messaging and objectives
that vary from one campaign
to another, an essential step
towards improving attribution
models is to consolidate
good data. “ We first linked
together all our locations
(websites, Facebook pages,
YouTube channel, advertising
server, etc.) in order to
consolidate the information,
identify our customers, and
follow their pathways more
easily. We wanted to get a
more comprehensive picture
of our users’ behaviour
within our digital ecosystem.
Ultimately, two things allow
us to analyse purchasing
behaviour: the digital loyalty
card and MyDermacenter.
These platforms give us total
visibility for a small portion of
our customers: the most loyal
and committed. They act as
“ samples ” to build models and
glean information on the rest of
the customer base as well as on
prospective customers ”.
... analysing it
By dissecting these
consolidated performance
indicators, Vichy can review
its biases. However, the brand
wants to go beyond managing
its return on digital investments
by extending this process to
offline sales. This is a topic
the team is currently devoting
considerable energy to: what
are the best indicators for
“ brandformance ” with regards
to purchases made at points of
sale? “One shouldn’t see things
in terms of branding versus
performance. The aim of all
our campaigns ultimately is to
generate sales in pharmacies.
We are working on solutions
to link involvement on our
websites to conversions at sales
points. For example, how does
one attribute point to a visit on
the Store Locator compared to a
visit on a product page in terms
of their relative impact in-store?
The link to sales is very difficult
to model, but we are working
with research institutes and our
media partners to come up with
a solid methodology”.
... and democratise it
A final goal for Vincent Labasse:
putting into place the tools
to allow marketing teams
to manage their campaigns
more easily. “ We are starting
with simple initiatives to allow
teams to better understand the
effectiveness of each point of
contact. For example, we have
standardised and simplified
our dashboards, since a few
indicators are enough to
understand the impact of a
lever. One has to be pragmatic.
Performance indicators can
be very opaque, but marketing
teams and management really
want more information ”.
ROI is an important topic
for Vichy, and not just for
digital. The brand entrusted
an econometric study to a
specialised firm a few months
ago.
“ Last year we concentrated very
heavily on three ranges of Vichy
brand products. We studied
the impact in terms of sales
of all of the methods we used:
media, advertising expenses,
POS advertising expenses, PR
campaigns, sales force...We
put all this information forward,
going back three years. It was
a substantial piece of work. It
took a lot of time and energy
internally to leverage the raw
data required for the study.
We identified a considerable
number of aspects we wanted
to have studied. They are now
providing us with good advice
when we want to launch an
operation. We are considering
launching other studies like this
to learn even more ”.
With 12,000 sales points
- none of which it owns -
and the launch in 2015
of an e-commerce site
(MyDermacenter.com),
L’Oréal Cosmétique Active
France and its Vichy brand
face especially difficult
problems with attribution.
Where to begin? The
company has decided to
capitalise on its data and
to optimise its internal
resources.
F
or L’Oréal, the third-largest
advertiser worldwide, the
topic of media investment
optimisation is not new.
What is new, however, is the
ever-increasing number of
levers at its disposal (formats,
technologies, media, etc.)
and the great volume of
disorganised information it
must analyse to ensure that it is
making the right choices.
Another difficulty for the Vichy
brand: the very “ fragmented ”
distributor landscape:
pharmacies and drug stores.
The majority of the brand’s
sales occur through retailers
and online retailers. “ This
network makes managing
the ROI more difficult ”,
says Vincent Labasse, head
of Digital. “ How can one
effectively manage the return
on the investment of each
channel without exhaustive
and reliable information on our
offline sales? Nonetheless, it’s
important not to fall into the
trap of “above-ground” digital
marketing: the profitability
of our investments is more
and more complicated to
understand, but it remains
essential for us all. Mastering
this data is a historical
opportunity for our brands to
put performance at the heart of
L’Oréal Cosmétique Active France includes five brands (La Roche-Posay, Vichy,
Sanoflore, Roger&Gallet and SkinCeuticals). The group launched “ MyDermacenter ”
in June 2015, which includes both an e-commerce website and a store in Paris.
“ MyDermacenter ” provides consumers with a personalised diagnosis, chatting with
advisors backed by a qualified pharmacist, a forum and more.
There are two main goals: acquiring data on clients and providing an online sales
model for traditional distributors like pharmacies and drug stores.
[Vincent Labasse is Head
of Digital Marketing for L’Oréal
Cosmétique Active France]
a complicated network
andconcreteanswers
L’Oréal Cosmétique Active
case study
9. spotlight
social networks are jealously
guarding their turf
17 // Chapter 1 // beyond “ last click ”16 // Chapter 1 // beyond “ last click ”
Social networks and their impressive amounts of data kindled hope among
advertisers, especially Facebook and Twitter, which regularly launch new
marketing tools. But are they sustainable solutions for brands?
T
ake the example of the un-
disputed leader in the field,
Facebook. According to an
article in Le Nouvel Économiste
published in April last year,
“ Facebook claims to have close
to 1.6 billion monthly users in
its social network. Approxima-
tely 1 billion people - a third of
Internet users worldwide - use
it every day ”.
The group cashes in on
19 % (35 % for Google) of the
70 billion dollars spent on mo-
bile advertising worldwide in
2015, according to the consul-
ting firm eMarketer.
Facebook is careful to maintain
its dominance in the field, pur-
chasing Instagram in 2012 and
WhatsApp in 2014. “ Facebook
is promoting Messenger and
WhatsApp as services through
which customers can buy pro-
ducts and interact privately with
companies ”, writes Le Nouvel
Économiste.
“ For example, the Dutch avia-
tion firm KLM provides e-tickets
and boarding cards by Mes-
senger. It can also be used to
chat with customer service. It
is already possible to request
an Uber car by Messenger,
without having to open up the
company’s app. [...] Facebook
has already become a kind
of ‘universal passport’ on the
Internet. ”
Blind trust
The journalist Will Oremus,
writing in Slate in 2012 on Face-
book’s purchase of Instagram,
highlighted that “ Surely Face-
book, with its 3,000-plus em-
ployees worldwide, could have
just built something similar to,
if not better than, Instagram for
a lot less money. So why did it
pay $1 billion?
Because for Facebook, this
isn’t just about improving its
photo-sharing app. It’s about
domination. Facebook doesn’t
want to be one among a num-
ber of options for sharing your
personal content with friends
and strangers. It wants to be
the only option. Just as Google
rakes in revenue by ruling
search, Facebook’s business
model depends on monopoli-
zing sharing on the Web ”.
Therefore it should come as
no surprise that the company
is unwilling to provide access
to its raw materials – its data.
“ Facebook states that it ano-
nymises your data when it sells
it to third parties ,” wrote Will
Oremus in an article published
in 2014. “ The company insists
that it anonymizes all your data
in its dealings with third-party
advertisers, so that they can’t
connect your behavior and de-
mographic information to things
like your actual name or email
address. And it apparently
expects you to trust its own
security mechanisms implicitly.
Its attitude seems to be that
you shouldn’t worry about the
giant file of information that
Facebook is collecting on you,
because Facebook will only
use that information for benign
purposes, like showing you
more relevant ads. ”
In this atmosphere of secrecy
and fear, Facebook is improving
and refining its data by ma-
king its algorithm increasingly
sophisticated. It launched five
new buttons in February 2016:
“ Love ”, “ Haha ”, “ Wow ”, “ Sad ”
and “ Angry ”. This would appear
to be the most efficient way to
increase its “customer intelli-
gence”: asking them directly.
However, these users “ have
begun to write increasingly
fewer status updates. This
has reached the point where
it has become troublesome
for Facebook ”, the newspaper
Les Échos wrote last April.
“ According to The Information,
the social network has put into
place a working group at its
London offices dedicated to fin-
ding ways to persuade users to
share more personal informa-
tion. One of the first initiatives
implemented by this team was
the “On this day” notifications,
which invites you to share a
photo, status or personal de-
tail, published on the same day
several years ago”.
Other tools have
been announced but not confir-
med, such as allowing users to
earn money from their posts
through a tip system (“ Facebook
considers letting users add a tip
jar to make money from posts ”,
theverge.com, April 2016).
Will social networks open up
or clam up to advertisers? For
Arnaud Lauga (Performics), this
is a question of “ turf ”: “ None or
almost none of the information
can be collected or used outside
of its own environment. The only
actions you can monitor and in-
clude in broader attribution
studies are clicks co-
ming from “ social
ads ”. The three
main networks,
Facebook,
Twitter, and
to a lesser
extent
Insta-
gram, are
“ walled
gardens ”.
The rich
data they
collect can
only be used
in their own
environment:
when you buy a
campaign, you are
told how many views
and clicks you got, but that’s it,
you don’t get any information for
each cookie or user ID. In short,
you get nothing. It’s the same
story with Amazon. However, I
think these “ walled gardens ”
will open up one by one in the
medium or long term, for fear
of losing business. For example,
Google’s DMP (Audience Mana-
ger) will open up to DSPs other
than DSM ”.
When Twitter announced that
it would significantly increase
the number of characters in its
messages, this was not just in
response to the requirements of
its increasing number of users,
but it was also to ensure that
it could host the longer texts
which its members currently
link to outside websites. This is
being done to keep things “ in
Twitter’s own garden ”, as Will
Oremus points out. “ Twitter
isn’t raising the character limit,
it’s becoming a walled garden ”.
It would therefore seem that
brands shouldn’t have high ex-
pectations in the medium term
for social networks. All you can
do is to learn to use them bet-
ter, optimise the free levers and
dedicate more time and human
resources to them, while kee-
ping branding at the forefront.
“ Only they can put their rich seams of data
to use: when you buy a campaign, you don’t get
any information for each cookie or user ID.
In short, you get nothing ”.
10. 19 // chapter 2 // bringing in the offline world
Ideally, attribution/contribution should not be limited to the digital realm,
no matter how rich it is. But are there any bridges between the two worlds?
First solutions are beginning to appear. Zoom to the “ Offline ”.
bringinginthe
offlineworld
18 // chapter 2 // bringing in the offline world
chapter 2
“T
here is a longstanding
culture of measurement
in the offline world that
is solidly rooted in existing
practices. However, it isn’t
about measuring performance,
it’s about measuring the target
audience. You buy exposure, not
conversions ”, explains Laure
Debos, Analytics, Research &
Insights Director at Publicis
Media, on the main difference
between Online and Offline.
On the rare occasions when
you purchase performance
in Offline, you link the exact
timing of a TV spot to the
number of calls to a Call
Centre or the number of
connections to a website, for
example. But this remains fairly
uncommon. “ Advertisers rarely
use TV solely for immediate
conversions, they put more
of an emphasis on branding.
Generally, with Offline, the line
between branding and impact
on sales is much more vague
than in Online, where we draw
a sharp distinction between the
two ”, says Laure Debos.
Different calendars and units
of measurement
The main unit of measurement
in Offline, and in TV in
particular, remains the GRP:
the number of individuals within
a predefined segment, such as
the famous “ housewife under
the age of 50 ”.
“ In digital, we measure certain
indicators. This is another
difference between the two
worlds ”.
As for remuneration, in Offline
you pay “ in relation to the
numbers of viewers reached ”,
says Laure Debos. The GRP
cost is negotiated with the
media company, and there
is no need for an attribution
model since a single institute
(Médiamétrie) delivers the
results and states how the GRP
was broken down.
Médiamétrie is a certifier, the
“ control tower ” for TV and
radio campaign results. The
equivalent for paper media is
the ACPM (Alliance pour les
Chiffres de la Presse et des
Médias).
“ Médiamétrie is financed by
the ecosystem as a whole (TV
channels, agencies, etc.), which
guarantees its impartiality.
There is no equivalent in the
digital world ”. This is a real
problem, according to Laure
Debos: “ Our online partners
and service providers have a
conflict of interest. They provide
results for campaigns that they
themselves have managed.
Many of us are trying to get
Médiamétrie to have the same
kind of role in digital as it has
in TV. The institute has many
ongoing projects, but there
are even more sources of data
available in the market ”.
At the end of a TV campaign,
Médiamétrie provides the exact
GRP out of the target provided
by each channel. Invoicing
occurs afterwards, and in
proportion to these results.
“ You often hear that digital is
more accessible, but in the
Offline world, you only pay
for what you get ”, says Laure
Debos.
Another major difference
between Online and Offline
is “ time-frame windows ”.
The common theme strategy
and the “ One Shot ” strategy
are both possible in Online
and Offline. However, the
implementation timeframe
is very different: Offline
often requires advance
planning (with regards to
openings in advertising
networks), and offers only
infrequent measurement
opportunities - 2 to 6 times
per year, according to media
companies. Modifying a
campaign is difficult. In
contrast, Online often allows for
daily measurement and easy
changes.
“ The current trend is to
measure how radio and TV
are drawing customers to
the website ”, says Laure
Debos. “ There are many new
technological solutions. The
tools provide numbers on
visits linked to TV campaigns,
often during the five minutes
following the spot ”.
Alexis Mollet (La Poste) used
one of these tools in his latest
TV campaign. “ In the long term,
the goal is to understand how
people reached by the spot
behave in the digital world. Is
the conversion rate better?
Is the average shopping cart
bigger? ”
Most members of the EBG
Task Force are very eager to
measure the effectiveness of
their TV campaigns through
these new solutions. Some
have tried even more radical
solutions: suspending TV
campaigns entirely for a year,
in order to see if anything
changes!
“ There are many other ways of
going about this ”, says Laure
Debos. “ For example, one can
link a consumer panel to a
media panel, to see if families
exposed to an offline
thenumbers
94 % The percentage of conversions
that occur in the physical world.
Source:Facebook / Atlas
“ For TV, we pay in terms of the viewership reached.
No one needs an attribution model since a single
research institute (Médiamétrie) tells us how
the GRP was divided ”.
lll
11. 21 // chapter 2 // bringing in the offline world20 // chapter 2 // bringing in the offline world
lll
“ You might wonder why the market has waited so long (around 50 years)
to take interest in such an old and reliable method such as econometrics.
The reasons are technical. We hear a lot about the development of the Internet,
but not about developments in computing power and the statistical suites
that have become broadly accessible in turn ”.
[ Laure Debos, Analytics, Research & Insights Director at Publicis Media ]
campaign make more
purchases than a family that
hasn’t been exposed. But
ideally - and this would be
revolutionary - we would like
to be able to measure and
quantify the impact of branding
on the performance of digital
campaigns”. TV campaigns
have long-lasting effects thanks
to persistence (the memory it
creates) and repetition.
Focus on econometrics
Econometrics is an ancient
science that has been much-
discussed in the past ten years.
What exactly is it? Statistics and
mathematics: “ Econometrics
is the analysis of time series ”,
explains Laure Debos. “ The
idea is to observe variations
over a very long period. You
determine what you want to
model; for example, variations
in a manufacturer’s sales. You
then go through two years of
sales data, week by week, and
look at anything that could have
influenced the curve: price
changes, publicity campaigns,
weather, school holidays, media
and marketing investment. You
get two years’ worth of data for
each of these, with a weekly
value for each. All this data is
processed and analysed by a
statistician who tries to account
for changes in sales. The goal
is to find a model that “ sticks ”
closest to the real data ”.
Once this substantial piece of
work is complete, econometrics
allows you to “ play ” with the
model, for example, in order
to determine at what level of
investment campaigns begin
to be profitable, and at what
level of investment they stop
yielding returns. Laure Debos:
“ Typically, a digital campaign
produces results very quickly for
a “ minor ” investment, but if one
invests too much, the saturation
point is reached very quickly ”.
For their part, TV campaigns
require more investment,
but models often show that
the saturation point is rarely
reached.
The model can also calculate
how long the impact of the
campaign lasts (persistence):
this helps you understand
how to spread out marketing
actions.
You might wonder why the
market has waited so long
(around 50 years) to take
interest in such an old and
reliable method. Laura Debos
believes that the reasons
are technical: “ We hear a
lot about the development of
the Internet, but not about
developments in computing
power and the statistical
software that has become
broadly accessible in turn. Ten
or 15 years ago, we didn’t have
the computers required to run
the complicated programs
required. You hear about it
these days because it’s like a
contribution model: you need
to look three months into the
past for digital work, compared
to two years in econometrics.
The only difference is that the
contribution model is used to
remunerate a partner, while in
the offline sphere, partners are
already remunerated ”.
It is an expensive method since
it requires considerable human
expertise as well as “ looking
beneath the bonnet ”. It also
requires the client to be entirely
on-board and to provide all the
necessary data, some of which
can be sensitive.
The still-tenuous link
between Online and Offline
Where are we in terms of
the integration of Offline into
attribution/contribution? How
can we manage the impact on
measurements? What kind of
progress can we hope for in the
medium or long-term?
For now, all we really know how
to do is measure attribution/
contribution with or without
a parallel offline campaign.
For example, we can observe
that the contribution of SEO
increases during a TV campaign.
In these cases, Offline is more so
contextual information,
“ Econometrics allows you to ‘play’ with the model
to determine at what level of investment campaigns
begin to be profitable and at what level of investment
they stop yielding returns. Typically, a digital
campaign produces results very quickly for
a “ minor ” investment, but if one invests too much,
the saturation point is reached very quickly ”.
lll
“One can link a consumer panel to a media panel, to see
if families exposed to an offline campaign make more
purchases than a family that hasn’t been exposed.
But ideally – and this would be revolutionary – we would
like to be able to measure and quantify the impact
of branding on the performance of digital campaigns”.
special report
12. 23 // chapter 2 // bringing in the offline world22 // chapter 2 // bringing in the offline world
rather than an integral
part of the model.
“ You would need to be able
to add cookies to TV ”
“ To manage the impact on
measurements, you would
need customers who see a TV
campaign to produce cookies ”,
says Laure Debos. “ You would
need “ TV cookies ”. I give the
example of TV because that’s
the medium where we’ll reach
that point first, because that’s
where the most information is
available. You would need to
link a cookie to “ people-based ”
marketing, as Facebook puts it,
by asking customers directly or
by merging information ”.
Reconciling Online and Offline
is a vital issue. “ The frantic
pace of measurement in Online
is still a bit crude. We are far
from perfecting our techniques,
and the market is in the process
of trying everything out: what
data is best, which partners to
choose...? The lack of precision
is a bit shocking to those who
come from the rigorous culture
of the Offline world. On the
other hand, we have a lot to
learn from Online, in terms of
reaction speeds, flexibility and
the test & learn approach ”.
We are starting to see solutions
based on the punching power of
social networks in combination
with tools like Atlas that can
reduce the distance between On
and Offline (described opposite).
From Web to store
and vice-versa
“ In incidents of fire, accidents
and miscellaneous risks
zoom
Atlas and DoubleClick: the first tools for a reconciliation
Atlas and
DoubleClick are
comparable tools,
in the sense that
they are “people
based” instead of
“cookie based”.
Cookies are linked
to a computer,
browser or device.
A cookie on a PC
can therefore
be connected to two people in the same household. Atlas and
DoubleClick work on the same identifier: the fact that a person
is “logged in”. One same email address can correspond, for
example, both to a loyalty card at a physical sales point and to a
username on a social media network. The two can be linked.
The Atlas solution, purchased by Facebook from Microsoft in
2013, uses clients’ Facebook usernames to connect the link
between Online and Offline: “ All the information is anonymised
in order to comply with CNIL ”, says the Performics team. “ Last
year, one of our clients, a bank, was able to measure that 67 %
of its impressions were delivered to its target age group of 13-
24 years old. But we were also able to discover that the campaign
was even more effective for 25 to 34-year-olds ”!
As for Google, it has launched its own solution integrated into
DoubleClick Manager (DCM), which works within the Android
environment, Gmail addresses, etc. According to the Performics
team, “ They have a little less coverage, but the tool is very
reliable from a statistical point of view ”.
Another solution is being reviewed by the members of the EBG
Task Force: “Call Tracking”. Specifically, each advertisement
and advertising channel are given their own phone number,
generating phone calls to the call centre. This is invisible to
clients, and the call centre keeps a record of the numbers called.
This subsequently allows you to measure the profitability of each
channel.
The main limitation of Call Tracking is determining when a sale
is actually closed: “ When do you deem a sale is made? Is a lead
enough? After all, Call Tracking measures leads and not actual
conversions ”.
However, advertisers face a dilemma. They aren’t always ready to
provide certain information to third parties, even if this is necessary
to produce better analysis. That’s why Sony Mobile, for example,
prefers to use internal tools to create its models (read p. 12).
within the insurance sector
(automotive and residential),
80 % of the pre-purchase phase
is conducted digitally, and
much less is being done in the
physical world. We therefore
need to guide the flow towards
sales points ”, says Rodolphe
Pachot (Allianz France).
Meilleurtaux.com, a pure player
with more than 200 sales
points, has the same problem
(see next page).
If digital campaigns can
complement in-store physical
traffic, the opposite can also
be true: advertisers relying
on offline campaigns to bring
traffic to their websites. This
is clearly the case for Meetic,
whose TV campaigns and
display ads bring new members
to its website. Aramisauto.
com, a pure player, goes a
step further: it uses offline TV
campaigns to bring prospective
clients to its website, knowing
that the website will then bring
clients to physical points of sale
to finalise the purchase.
lll
thenumbers
187 In 2003,consumers could be
exposed to 37 different contact
points.“ Twelve years later,in 2015,
our studies found 187 potential points
of contact between the brand and the
consumer ”,says Laure Debos.
Source:ProprietaryTouchpoints
study - Publicis Media.
special report
13. 25 // chapter 2 // bringing in the offline world24 // chapter 2 // bringing in the offline world
The system can distinguish
visits linked to the spot from
normal traffic. “ We activated
the system well before the
campaign started so that it
could observe normal traffic on
the website ”.
Indirect post-view visits outside
the six-minute timeframe are
also estimated by excluding
natural visits (SEO, direct
access) or AdWords. This allows
us to better estimate the “ halo ”
effect of the TV campaign.
“ Without information and
measurements, we can’t
improve and we can’t optimise
our actions ”, says Guillaume
Gozlan. “ We now have access to
data on the type of product our
website users are interested in:
real estate loans, automobile
loans…The TV spot focuses
on real estate loans, but it
ends with the following words:
“ Real estate, automobile,
consumer loan, professional
loan, meilleurtaux.com – may
the best win ”. The idea was to
give information on the different
sectors we work in. We were
able to observe that from the
very beginning of the campaign
the visits were spread across
different sectors ”.
A good analytical lens
With a 16 second TV spot, shot
in 3D and aired on multiple
TNT and cable channels, linked
to transmission sponsorships
on M6 (Capital, Zone interdite,
Recherche appartement
ou maison, series, etc.),
Meilleurtaux.com now has
better data to help guide its
future campaigns. “ This is
a strong analytical lens that
helps guide our strategy.
We’re looking for the best
compromise to optimise
our budget. This pragmatic
approach comes from the
Internet. I’m also interested in
the distinction between mobile
devices and desktops. There
are now many multitaskers
in France, consumers who
watch TV while surfing on
their tablet ”. Beyond its TV
campaigns, Meilleurtaux.com
uses local levers (posters,
exhibitions, etc.) determined by
local conditions.
Meilleurtaux.com, a pure
player with 230 franchises
in France, has been on TV
since 2011, first in typical
TV spots and then as a
show sponsor.
Guillaume Gozlan chose
both options for 2016.
Above all, he chose a
measurement solution.
Objectives: quantify the
increase in traffic directly
linked to TV and measure
the benefits in terms of
conversions.
“W
e have had an
“ Analytics ” culture
for a long time, and
we’re happy that new tools are
appearing to allow us to better
understand the impact of our
TV campaigns on our website ”,
says Guillaume Gozlan,
Marketing and IT Director at
Meilleurtaux.com.
“ We have chosen Realytics,
which offers an advanced
technological solution,
developed by a French
company. Our goal is simply
to optimise our media plan: to
what extent do TV campaigns
generate website visits? How
many estimates, and on what
products? At what times and
on what channels do we get the
best results? ”
Behind the scenes, the system
they put in place starts by
detecting the spots and show
sponsorships. “ We’re basically
talking about robots watching
TV ”, says Guillaume Gozlan.
“ And yes, we’re very confident
that they recognize our spots.
And since the tool is linked to
our website, it can show us
peak visit times or specific
visits during a window of six
minutes after the spot is aired.
This applies both to show
sponsorships and TV ads. ”
And it does this with subtlety.
[Guillaume Gozlan, Marketing
and IT Director.]
“ measuringtheimpact
ofourTVspots ”
Meilleurtaux.com
“ We analyse site traffic during the six minutes
after the spot is aired. Indirect post-view visits
outside the six minute timeframe are also estimated
by excluding natural visits (SEO, direct access)
or AdWords ”.
case study
14. 27 // chapter 2 // bringing in the offline world26 // chapter 2 // bringing in the offline world
keypoints
channels, especially in radio,
and more occasionally on TV.
We also have poster campaigns
and our service centres are key
communication points ”.
A project in three phases
On the digital side of things,
Euromaster focused for a long
time on Search. “ We work in
an essential needs market:
tyre replacement, oil change,
brake pads...Our prospective
customers are therefore
customers who intend to make
a purchase, and we have put
a lot of emphasis on active
research channels (SEM,
retargeting), which give us the
best return on investment ”.
Nowadays, the company is using
many other levers. Therefore
we need to know with some
precision the effectiveness
of each channel and each
campaign. This helps us spend
our money more wisely, and
of course, to sell more! “ We
know that some levers are
over-emphasised, and others
under-emphasised. If we know
what part of our budget is
most effective, we can saturate
it before moving on to other
channels ”.
Their attribution/contribution
project was launched in 2016.
Romain Perrin built it in three
phases, starting by analysing
the deployment of various
digital channels.
This will be followed by a
cross-device stage, and then
a cross-channel phase using
econometrics amongst other
things.
“ The technology has a head
start on the marketers ”.
So far, Romain Perrin has
completed a study of the tools
currently available on the
market. “ Software solutions
are currently very effective.
Often they are even a little
bit ahead of the advertisers.
We are in the midst of the
virtuous cycle. Technological
advancement makes us ask
new questions, which in turn
improves these tools and
helps them become more
sophisticated. I have sourced
a solution that corresponds
to Euromaster’s current and
future needs. It’s not the ‘luxury’
model, but it is accessible and
comprehensible for our traffic
managers in various countries,
which is essential. Behind every
attribution campaign are human
resources. You need the right
people to analyse the data you
collect ”.
The software, installed in
May 2016, will soon allow
Euromaster’s digital strategy
to evolve. “ The ROPO (research
Online, purchase Offline) effect
is very important for us. These
are not impulsive purchases,
and the consumer conducts
careful research before making
a decision ”.
Euromaster is a click
and mortar company
operating in an essential
goods market, automobile
maintenance. The digital
world allows the company
to direct its customers to
its service centres. Having
long depended on Search,
Euromaster is now playing
with several other levers. It
has recently equipped itself
with a software attribution/
contribution solution in
order to better identify the
role of each channel.
“T
he historical heart of
our business is the
tyre ”, says Romain
Perrin, Digital Acquisition
& Retention Manager at
Euromaster. “ Several years
ago we expanded our service
offerings to include ongoing
maintenance ”.
Euromaster operates in
17 countries and has a network
of around 2,300 service centres
in Europe. For this company,
the Internet serves as a lever
for generating traffic in its
service centres. The level of
online interactions varies.
It includes transactions
(e-commerce), appointments
(individuals and commercial),
estimates, service centre
locator and downloadable
coupons for services.
Identifying consumers acting
with the intent to purchase
“ We leave it up to consumers
to choose the entrance point
that suits them ”, says Romain
Perrin. “ This Web to Store
strategy has been in place for
several years and from now
on, I’m looking to emphasise
the cross-channel aspect. We
have a lot of presence in offline
“ In terms of attribution/contribution, you should not set out to immediately find the Holy
Grail ”, says Romain Perrin. “ It’s better to proceed calmly and to understand exactly
what’s happening, step by step, between the digital and physical worlds. As is often the
case in the digital world, one often needs to test things before deploying them across the
board. Even if one attribution tool suggests that a particular channel is a ‘gold mine’, I’m
not going to divert all of my budget to this channel. There is no 100 % accurate model.
The model can change depending on the date, the hour, the device, the market... Finally,
the digital and communication teams need to work hand-in-hand in-house. The data
deployed for the study, just like the data resulting from the study, is cross-sectional ”.
[Romain Perrin is Digital
Acquisition & Retention Manager.]
“ thedigitalworld
workingfor
thephysicalnetwork ”
Euromaster
zoom
The figures
• Euromaster (Michelin group) has around 11,000 employees
in Europe and 1.8 billion euros in sales.
• Euromaster sells 7.8 million tyres each year, including
2.5 million in France.
• Euromaster France was named best chain store in France in
2015-2016 in the automotive sector.
case study
15. spotlight
29 // chapter 2 // bringing in the offline world28 // chapter 2 // bringing in the offline world
The market is buzzing with solutions to better exploit first party data, and
you can even try your luck with second and third party data. However,
marketers are being more cautious with “ full DMP ”: it’s better to do
things one step at a time.
T
he DMP (Data Management
Platform) concept has
provoked debate. “ DMP
has been a commonly-heard
buzzword for the past two
years ”, says Arnaud Lauga
(Performics). “ Many compa-
nies want to quickly build up
their DMP and to acquire a
full toolset. However, the per-
ception has changed this year.
Marketers aren’t throwing
themselves anymore into these
very demanding projects. They
understand the need to proceed
step-by-step, and that it’s better
to start by using the tools they
already have ”.
And here they are in a strong
position! So-called ‘first party’
data is the most qualitative.
Advertisers collect this data
through their own interactions
with clients and prospective
clients. Exposure to media
campaigns, site pathways,
purchases/conversions and
CRM information in general. It
exists only within the company.
“ However, there is still work to
be done to optimise first party
data! ” says Arnaud Lauga.
“ It’s the most valuable and
inexpensive data. Of course,
it requires energy and an
investment in human resources,
but it’s worth working on it
before spending 400,000 or
800,000 euros for a full DMP ”.
After all, by collecting and
merging first party data and then
processing, dividing and then
using it, it is already a DMP - a
“ first party DMP ”! After that, one
can collect more data through
advertiser partnerships or editors
(second party data). Finally, third
party data is much less qualitative,
but also comes in great quantity.
However, for most advertisers,
proceeding step-by-step is the
key to success.
Some, like Yacine Mahfoufi,
France Marketing Director of
Alcatel-Lucent Enterprise who
interviewed in January 2016 for
emarketing.fr, point to the tool’s
weaknesses. “ The benefits
of DMPs are currently over
estimated for five reasons: the
inability to merge with external
data, the poor matching rates
between your data and the
CRM, an unproven ROI since
anonymised third party data
is expensive, lack of clarity on
privacy and under-appreciated
“ There is still work to be done to optimise First party
data! It’s the most valuable and inexpensive data.
It requires an investment in human resources,
but it’s worth working on it before spending
400,000 or 800,000 euros for a full DMP ”.
DMPandCRM:towardsamore
pragmaticapproach
technical complexity -
implementing the tag plan
with the matrix that links your
offers on the one hand and
messages modified for email
and display on the other hand
is a considerable piece of work
that must be completed before
making a start ”.
Linking a cookie to each
element of the CRM database
Among the more interesting
recent technological innovations
is an improvement of client
relationship management
thanks to “ CRM onboarding ”.
In simple terms: “ This is
about linking digital identifiers
(cookies) to each piece of
information in a CRM database.
Has someone logged on? Link
a cookie to them. Is someone
sending an email? Similarly,
assign them a cookie. And so on
and so forth ”. This data must be
anonymised by encryption.
As of May 2016, there are four
service providers in the French
market.
In terms of the downsides of
CRM Onboarding,
aside from high pricing that
can jeopardise ROI, there are
problems with the freshness of
the data. “ In the travel sector,
for example, a three-month-old
cookie can already be obsolete ”.
And the percentage of the
CRM database that can actually
be “ onboarded ” is a topic that
is up for debate. Achieving 25 %,
it seems, would be quite an
accomplishment.
“ CRM On-boarding raises
the problem of data
freshness. In the travel
industry, a three-
month-old cookie can
already be obsolete ”.
16. Cross-channel, and to be more precise, cross-channel devices, will change
everything.
excelling
atmultichannel
chapter 3
31 // chapter 3 // excelling at multichannel30 // chapter 3 // excelling at multichannel
“A
t present, when you
think you’ve reached
100 people, you have only
reached 42 at most. And when
you think that a consumer has
seen your ad 10 times, in reality
they have seen it 24 times. We
both overestimate target cove-
rage and underestimate repeti-
tion”. These statistics, provided
by Performics, show how diffi-
cult circulation and campaign
measurement are from a cross-
channel perspective.
The cross-device and
the dream of a unique ID
The cross-device does not
simplify matters - 40 % of
consumers today start a
purchase on one device
(computer, tablet, phone,
etc.), then finish it on another.
Even if today, the majority of
cross-device conversion re-
attributions take place between
two computers (office vs. home),
Arnaud Lauga (Performics)
reminds us: “ One person is
four cookies! One computer
at home, one computer at the
office, a smartphone, a tablet,
a laptop...The goal is to get
a unified understanding of
these four cookies by grouping
them under one ID. After that,
of course, one must make a
connection with the CRM, in
order to know that a particular
ID belongs to a particular client
- anonymously, of course. What
interests us is the information
we have on the client (RFM
segment, age, purchase history,
type...) and not their name ”!
Is this possible, in technical
terms? We need to turn back
to Atlas and DoubleClick (read
page 22). “ By using a people-
based perspective - the Atlas
solution - one of our clients
was able to observe that its
conversion rate increased by
29% when at least two devices
were used, compared to only a
desktop ”, says Arnaud Lauga.
“ Another advertiser in the
fashion sector discovered that,
according to Atlas, 13 % of
its in-store sales followed at
least one exposure to an online
campaign ”. These statistics are
from January 2016.
“ Another significant result
is that by using Google’s
DoubleClick solution with
cross-device measurement, a
communications company was
able to better re-attribute its
conversions, leading to a 110 %
increase in attribution to its paid
levers. Exposure on the mobile
device followed by conversion
through direct access on a
desktop is attributed by default
to normal traffic ”.
There is a third solution, “ co-
opt ”, where advertisers share
their logged data, for example,
Adobe, Signal, etc.
“ For all of these ”, says Arnaud
Lauga, “the bridge is the email
address, username or Facebook
ID. Both Facebook and Google
have enormous reach, so it was
only logical they were the first to
throw themselves into these kinds
of tools. But there are others like
BlaBlaCar and LinkedIn that also
have considerable reach and are
in a position to launch their own
solutions ”.
zoom
Mobile continues to buzz
“ Online shopping on
mobile devices doubled
in 2015 ”, according to
the magazine L’Usine
digitale (March 2016).
A study carried out
in January 2016 by
the Centre for Retail
Research in eight
European countries
(France, Germany, Spain,
Italy, Netherlands,
Poland, UK and Sweden),
as well as in North
America, found that
“ online sales have
reached 36.5 billion euros, an increase of 17 % in a single
year. This puts France in third-place in the EU, after Great
Britain and Germany. This strong overall result hides changes
in the types of device used to make these purchases, with an
increasing reliance on smartphones.
For the time being, though, a majority of purchases are made
by computer. Online shopping doubled in 2015, reaching
7 billion euros, compared to 3.7 billion one year earlier. In
other words, each French mobile device customer spent
282.85 euros in 2014, 521.77 in 2015 and will likely spend
711.67 in 2016.
This is in part due to the fact that “ one of the factors causing
this growth is the increasing rate of smartphone usage ”,
according to one of the authors of the study, noting that at
58 %, this rate is 4% above the European average.
Smartphone purchases tend to be small, with an average
shopping cart of 49.74 euros in 2015. It will decrease in 2016 to
47.54 euros ”.
This was highlighted by the members of the EBG Task Force:
“ We have observed the increase in mobile sales as well as
the difficulty at this stage in doing cross-device tracking.
Companies such as Facebook are proposing to use their
logging across multiple devices in order to reconcile these
two. At this stage, not many can use this, and it represents a
pretty small sample size ”.
17. variable into the equation
because if you only look at the
usual 30 day attribution window,
there is a risk you might miss
something important ”.
The members of the EBG 2016
Task Force made a list of the
various A/B test methodologies
and their advantages and
disadvantages. They can be
divided according to three main
criteria:
- Before/after split
- Geographical split
- Random audience split
The last type is the most
reliable, without any seasonal
or geographical bias, but it is
impossible to use with certain
levers (SEA, affiliation, typical
display).
Dividing by before/after
introduces a major seasonal
bias and makes it difficult to
isolate two different groups.
Dividing geographically (one can
divide France into 2, 8, 10, etc.
regions) introduces an obvious
bias and implementation can
be complicated for display and
affiliation.
Incremental performance
Trading desk RadiumOne
has implemented “ in-house ”
tools, which it first applied to a
display campaign. An A/B test
was carried out to measure
the advertising uplift (the
incremental performance): “ We
made use of the binary nature
of RTB - it’s either a win or a
loss. We found a 54 % increase
among those who were exposed
to the campaign compared to
those who were not. Of course,
one has to take into account
that there is a bias inherent in
this method ”.
We then tested the campaign
strategy – what would happen if
we made incremental changes?
“ We observed that the
conversion rate increased very
quickly up until four exposures,
and continued to increase but
at a lesser rate until seven
exposures, but that any further
exposures made the conversion
rate drop ”!
Third test: the trading desk
used client data - geolocation
of the mobile device users at
the points of sale. “ Using GPS
coordinates with an accuracy
of 15 to 20 meters, we were
able to see if users who are
exposed to the campaigns
went to the stores. This is often
used if a company sends out a
notification that it is holding a
sale on a particular day ”.
A
final test incorporated CRM
data from a company and
measured sales compared to
advertising exposure: “ The
more companies digitise their
CRM databases, the more
we can use these kinds of
techniques ”.
The Performics team then
presented a client case study
carried out using random
groupings. “ We divided users
of an e-commerce website
into three groups and a
control sample. We did not
use any retargeting on the
control sample. Before
“I
ncremental management is
very important for us, since
we believe it represents
the true added value provided
by an advertisement ”, explains
Arnaud Lauga (Performics).
“ It’s the ideal indicator, the Holy
Grail. Supposing an advertiser
closes 1,000 sales in a world
where they have advertising,
compared to 900 sales in a world
where they haven’t. They would
be pleased with this result. Our
goal is to use this indicator in an
even more sophisticated man-
ner: which advertising levers
performed best? ”
A/B testing and client cases
As in medicine, this
measurement is made in
comparison with a “ placebo ” -
a sample that was not exposed
to the advertising lever.
But if four levers (SEA, display,
retargeting, affiliation) are being
studied, for example, there
can be a great many possible
combinations depending on
whether one uses one, two,
three or four of the levers.
“ A/B tests have limitations -
in reality, it’s very difficult to
find a population that hasn’t
been exposed to any of the
levers – unless you lock them
up in a cave! No two cases
can be entirely isolated from
one another. And to measure
each lever, everything has
to be duplicated by as many
times as there are possible
combinations. It is complicated
or even impossible from a
technical standpoint, and it can
have a major impact on the
budget ”. Another problem is
that advertising doesn’t always
have an immediate impact.
Should one measure after one
day? A month? A year?
“ It is important to build this
32 // chapter 3 // excelling at multichannel 33 // chapter 3 // excelling at multichannel
thenumbers
12 % 12 % of conversions are not
captured - in other words, the
client finishes their conversion
without it being attributed to
a marketing action, since they
changed devices in the meantime.
(Sources:Atlas internal data,March
2015 & Nielsen Digital Ad Ratings
industry norms for US,July 2014)
understandingthevalueadded
ofeachchannelIncremental management allows advertisers to determine the contribution
of each lever in comparison to an initial situation in which this lever was
not activated.
“ A/B tests have limitations - in reality, it’s very
difficult to find a population that hasn’t been exposed
to anything - TV, SEA, affiliation - unless you lock
them up in a cave! And to measure each lever,
everything has to be duplicated by as many times
as there are possible combinations. It is complicated
or even impossible from a technical standpoint ”.
lll
special report
18. 34 // chapter 3 // excelling at multichannel 35 // chapter 3 // excelling at multichannel
launching the campaign,
you have to ensure that the
same conversion rate appears
among the three groups and the
control sample. In this case, the
difference was minimal - 0.07 %
- suggesting that the random
division was successful ”.
Over the course of a month-
long campaign, Performics
measured an increase in the
conversion rate of 8.8 % on the
sales volume, (representing an
additional 10,000 sales).
“ It was clear that this was
a population that could be
retargeted, as it was composed of
visitors that the advertising client
had access to. Each retargeter
will make their own decisions
on how to reach out to the group
of clients they have access to,
concentrating for example on
‘ shopping-cart abandoners ’,
who are the most likely to be
converted, or on the visitors who
remained on the website for the
longest period of time ”.
You have to identify which
visitors are the most
susceptible to advertising. How
do they behave on the website?
Off the website? You need to
find some correlation and try to
target them.
The next step is to determine
the right level of pressure.
“ Multiplying the budget by
three doesn’t necessarily yield
proportional returns ”.
Identifying the maximum
budget for each lever
Which statistical concepts
can be applied to advertising
campaigns? Performics has
developed a mathematical
approach.
Members of the Task Force
were asked to work on a micro-
economic case study: “ If I have
10 euros, how many oranges
and chocolate bars should I buy
to maximise my pleasure? ”
In this case, “ decreasing
marginal utility” (the first
chocolate bar giving more
pleasure than the second
one, and so on) allows us to
determine that if the price
lll
Optimising the allocation of your budget across several levers
budget
With two levers
[Marginal CPA modelling in relation to
budget. The optimum level is reached
when the two levers have the same
marginal CPA.]
is equal, the choice between
oranges and chocolate should
be guided by maximising the
marginal pleasure produced
by each additional unit. This
leads to an equation: the
optimal allocation of the budget
between oranges and chocolate
is reached when the marginal
utility of an orange divided
by the price of the orange is
equal to the marginal utility of
chocolate divided by the price of
chocolate.
Assuming you haven’t given up
reading by now to find some
aspirin (and a chocolate bar),
it is worth noting that this
theory can be applied to the
media sector and allows one
to determine the optimal mix
of medias - sales or revenues
generated by channels, total
media budget to allocate
between channels, marginal
CPA or ROI.
“ It produces some indicators
we don’t normally see, such as
the ‘marginal CPA’. The optimal
budget allocation between
channels (which allows you
to maximise sales) is reached
when the marginal CPA of lever
A is equal to the marginal CPA
of lever B ”.
In another workshop, members
of the Task Force were invited
to dive into a table of figures
to answer the question: “ How
do you divide a budget of 7,000
euros between levers A and B in
order to maximise your sales? ”
With three or more levers,
creating a model becomes
essential (see graphs opposite).
For each budget level, the
model determines the optimal
allocation between levers A and
B to maximise sales.
This shows that the optimum
is not the same for each
overall budget level. The main
limitation of the process is that
it doesn’t take into account
secondary objectives such as
image. Moreover, the future
cannot be assumed to be
the same as the past, so the
calculations need to be kept
up-to-date.
Creating a model is essential
when you have more than
two levers
[Optimal allocation of a budget between
different levers. For example, for
830,000 euros, the optimal allocation is
52 % x 830,000 = 431,600 euros for the
SEA, and 11 % x 830,000 = 91,300 euros
for affiliation.]
special report
20. 39 // chapter 3 // excelling at multichannel38 // chapter 3 // excelling at multichannel
zoom
contribution questions, ” says
Simon Le Fur.
“ The first thing to do is to
deduplicate; you just can’t
work without it. After that,
you have to understand the
client’s purchasing dynamics.
At Greenweez, we are working
with food distribution profit
margins, so we base our
profitability on selling large
volumes. I implemented the
Eulerian Analytics solution.
They offer several attribution
models and, above all, allow you
to compare them. The last click
is the basic display, even if it is
not perfect. We compare it with
first click display, which gives
us an idea of where the client is
coming from. This information
disappears quickly, but it’s
good to find out what the client
did during the 30 or 90 days
preceding the purchase ”.
This helped Greenweez.com
identify major acquisition
channels that it has to be
careful not to cut off. “ For
us, it’s AdWords and price
comparators. If an AdWords
campaign costs too much, we
remember that analysis shows
that AdWords brings sales with
the first click and we optimise
the campaign so that we can go
forward with it ”.
On the other hand, Simon Le
Fur has been able to identify
“ declining ” channels, such as
affiliation. “ There are still many
companies that pay to have
promotional codes published
on sites other than their own.
Users will look for codes even
if they have already decided to
make the purchase. The last
lever in a session, before the
conversion, clearly shows this ”.
“ For me, ” says Simon Le
Fur, “ attribution should only
be used to take concrete
decisions. Eulerian lets us
allocate our budget. Everyone
in the company has access
to it, and the data is located
centrally so that everyone can
draw information from it that is
directly useful to the company.
This is a precondition for us
staying flexible ”.
Greenweez.com, founded
in 2008, has become
the “ Internet’s organic
supermarket ”. This pure
player has developed a
pragmatic approach to
attribution/contribution,
based on the Eulerian
solution.
.
I
f a company like Biocoop or
Naturalia has 5,000 to 6,000
listings, Greenweez displays
25,000. This was the initial stra-
tegy: hyper-choice. “ We wanted
to become the leading specialist
in organic sales on the Internet.
With revenues of 20 million
euros, that’s what we’ve done ”
says Simon Le Fur, General Ma-
nager. “ Our basic model is to
have a wide range of products,
at affordable prices. We sell not
just one but multiple types of a
single product, all organic ”.
Like any true pure player,
Greenweez learned very early
to listen closely to its clients:
“ They made it clear that
they wanted to do all of their
shopping online on the same
website. Today, we continue
to use feedback and are
working on an ambassador
program with Howtank. We are
connected to our clients, while
our physical competitors, who
have only partial coverage,
struggle to reach them. ”
Acquisition dynamic
The Greenweez team is made
up of only 25 employees. Each
marketing decision is therefore
taken in the most operational
manner possible. “ I used to
work at Sarenza, where I spent
a lot of time on attribution/
“ With 25 employees, we can’t go looking for individual IDs. But of course we are
interested in cross-device, ” says Simon Le Fur. “ For me, the only reliable indicator
is the percentage of transactions carried out on each device. We have identified a strong
appetite for making purchases by mobile device over the past year and a half.
We have to react to this and are working on improving the website to make it more
responsive. 15 % of transactions are carried out on mobile devices at present, with
a mobile device traffic at 25-35 %. Data like this is invaluable since it makes us aware
of deeper dynamics and lets us position ourselves correctly and get the timing right ”.
[Simon Le Fur is General Manager
of Greenweez.com]
“ identifying
value-creatingchannels ”
Greenweez.com
zoom
The figures
• Greenweez.com sells more than 3 million organic products
per year;
• Its clients are very loyal: they place, on average, five to ten
grocery orders per year;
• 500,000 individual visits per month;
• The average shopping cart is for 95-100 euros including
taxes.
“ We have defined our important acquisition channels,
which are AdWords and price comparators.
We cannot cut these off. ”
case study
21. 41 // chapter 3 // excelling at multichannel40 // chapter 3 // excelling at multichannel
oversimplifying things, you need
to go into the details of how
value is measured ”.
About the purchase cost
What really matters, at the end
of the day, is the purchase cost.
“ The idea is to find a purchase
cost for each lever, based
on the real link between a
contact point and a sale,
instead of dividing the total
number of sales by the cost of
the campaign, ” says Laurent
Bouten.
“ We haven’t achieved that yet. I
think it’s the most complicated
part, and it’s an important
question for everyone: should
you continuously carry out
auditing work in order to model
attribution, if you don’t use
it continuously? Everything
depends on the company’s
strategy and on its acquisition
context ”.
At Motoblouz, the current
priority is to internationalise:
“ We develop 99 % of this by
means of AdWords. Working
on attribution is therefore not
a priority. It’s an important
point for other companies that
do branding. But if you want to
start working back to the first
contact point, to go beyond your
own website and track all the
banners involved, it can get
expensive ”.
Affiliated and discount codes
Laurent believes that
outsourcing an attribution study
before putting things into place
to do it in-house can be a good
practice. “ The second step
would be to get this going at
a regular tempo that is suited
to the company’s means and
objectives ”.
He concludes that “ for us, we
haven’t done much reattribution
because we know that there
isn’t much crossover between
the channels. Originally, the
question of attribution started
with a debate on discount codes.
Many of my colleagues realised
that they were paying far too
much to certain discount code
providers. In reality, consumers
were going to Google to search
for discount codes having
already made the decision to
buy ”. (see the Compagnie des
Alpes case study on pages 36-
37).
“ It’s like with pilot fish - for
a long time we thought that
they were guiding marine
predators, when actually
they’re just taking advantage
of the slipstream! This led a lot
of advertisers to go so far as
equipping themselves with a full
attribution tool when it would
have been enough to use an
intermediate-level tool ”.
The pure player
Motoblouz.com, which sells
motorcyclist equipment,
has invested in attribution/
contribution since 2014,
while retaining, it must be
said, a sense of perspective.
The company already
uses two additional tools
to ensure that the data
collected is processed
correctly, but is careful
not to over-invest. The
company’s priorities push
it towards investing most
heavily in AdWords
“T
here are two important
aspects of attribution/
contribution, ” says Lau-
rent Bouten, Marketing Director
of Motoblouz.com. “ Measure-
ment first (understanding the
client pathway), strategy second
(deciding whether to act or
not). ”
Acquiring measurement tools
without producing analysis of
no interest: “ We have been
working with Eulerian for two
years to record our clients’
pathways. The tool records
all access to the site - where
the user arrived, what they did
when they clicked... At this point
there is no payment aspect, it is
pure measurement. ”
“ We complement this with
Mazeberry for data display.
This seems key to me, since
otherwise you are stuck with
a lot of data and it would be
a shame to only scratch the
surface. You can do your
own CPC, comparators or
retargeting analysis, but this
will lack granularity. ”
“ With CPC, for example, the
takeaways from a generic
keywords campaign will be very
different from the takeaways for
a shopping campaign. To avoid
[Laurent Bouten
is Marketing Director.]
“ avoidingtechnological
one-upmanship ”
Motoblouz.com
zoom
The figures
• 700,000 clients;
• 30 million euros
in sales;
• 11 years in business;
• 50,000 references;
• 17 million visits;
• 5 countries covered so far.
“ You can use a measurement tool to conduct
your own CPC, comparator or retargeting analysis,
but this will lack granularity. With CPC, for example,
the takeaways from a generic keywords campaign will
be very different from the takeaways for a shopping
campaign. To avoid oversimplifying things, you need
to go into the details of how value is measured ”.
case study
22. 43 // chapter 3 // excelling at multichannel42 // chapter 3 // excelling at multichannel
in practice
Élodie Hazac (Orange) gives
some advice drawn from her own
experience.
“W
hat do you have to take into account when
you are interested in attribution and you
aren’t a pure player? My opinion is that
internally, the most important thing is to have a
business plan to make sure it stays profitable.
If a few prerequisites are already in place, it’s
possible to have started to turn a profit on the
software in less than a quarter...after that, in
order to account for the integration of Offline in
attribution, there are solutions such as Google
Analytics once you have primary identifiers like
an email address, an order number or a client
number. Therefore, one must be prepared to
do quite a bit of work in advance to acquire this
unique identifier. If not, you pay for a tool that you
will use only partially.
It is also important to have the in-house
resources for processing the information
provided by the tool. The team that is putting it to
use needs to have a complete picture and be able
to act on the different visibility levers. This leads
you to the question of management of the project
- where should the team be located so that the
analysis is efficient, and who sponsors it?
Attribution is no more than an activity traffic
optimisation tool. Another question you might
want to ask yourself is: «What kind of relationship
should we put in place with our agency to deploy
the optimisations?»
There is a lot of excitement about these
attribution solutions, but a lot of time passes
between starting to think about it and the actual
implementation. These kinds of tools must fit
into a comprehensive digital overhaul project
that should include all the teams carrying out
«Paid», «Earned» or «Owned» campaigns. A
campaign can be effective, but if you bring traffic
to a website that isn’t optimised to clinch sales,
it creates a performance measurement bias:
you won’t know where to attribute sales! This is
what leads companies to get an attribution tool,
followed by a testing tool. One last point: as you
deploy the tool, make sure there is somewhere
for people using the tool or the information it
produces to share their ideas. Promote regular
exchanges of information».
“ Youincreasetraffic,
butifthewebsiteisn’t
optimisedtoclinchsales,
itcreatesaperformance
measurementbias ”.
in-house Organisation
and change management
This is a brand new field. Where to start? Advice from Julien Braun
(RadiumOne) and Arnaud Lauga (Performics).
#1
Choose tools suitable for your
business and global media
strategy. Adapt both the solution
and the measurement sample
to your needs. If you want to
improve online sales, measure
online sales! There are different
technologies behind different
measurements. Be clear about
your priorities before starting
#2
It is still very rare to find
advertisers using incremental
management, even though
this method of measurement
is more accurate than the
others. It is difficult to transition
from last click management
to incremental management.
Don’t rush through the steps.
Consolidate your data, unify it,
sort it and make sure you have
the support of the teams and
services that are involved.
#3
Change the way you think. “ One
of our clients optimised their
measurement by organising a
competition between agencies
(he used three) without
generating any additional
sales. In contrast, incremental
management is a very pragmatic
approach; it isn’t a philosophy.
It’s been used for years in
Offline. A/B test it, and you will
be convinced ”.
#4
Measuring is good, but you
have to then make use of the
findings. There has to be linkage
between measurement and
action. This can be done through
programmatic campaigns.
Get many people involved, and
make sure you have enough
resources in-house in order to
optimise your media purchase
strategies in accordance with
the lessons of your attribution
studies. You might come to
modify them pretty substantially.
“ Incremental management is a very pragmatic
approach; it isn’t a philosophy. It’s been used
for years in Offline. A/B test it, and you will
be convinced ”.
firststepstowards
incremental
measurement
our advice
23. our advice
Define and prioritise your needs
Take the time you need
to define what you’re
looking for, since
everything else will
follow. Then compare the
effort and costs to the
expected benefits.
Choose the appropriate
technologies.
In accordance with the
level of sophistication
required, your current
technology, and the
ability of your teams to
use and master them.
Go about it step by step
Take the time to build
your technological
architecture. Don’t wait
for the perfect ecosystem
to get started since the
environment is constantly
evolving. Get started,
44 // Multichannel attribution, contribution & optimisation 45 // Multichannel attribution, contribution & optimisation
even if you have only 80%
accuracy.
Don’t hesitate to carry out POCs
(proofs of concept)
And determine your
own benchmarks
(technologies, partners).
What matters is the
customised optimisation
of your business.
Challenge your attribution
models, while keeping
your current model as a
starting point.
Make medium-term projections
The system you implement
must be evolving and
unrestrictive. For example,
expect your online and
offline strategies to
become synchronised,
both in terms of marketing
actions and in-house
organisation (take the time
to communicate and get
the company on-board).
Draw from the lessons of the
market, including those
from other sectors.
Capitalise on your own data
It’s a gold mine! Take
advantage of it with third
parties that you trust.
People are key
Ensure positive
collaboration between
your partners.
getstarted
inasafeenvironment
By the Performics team.
“ Thesystemyouimplementmustbeevolvingand
unrestrictive.Forexample,expectyouronlineand
offlinestrategiestobecomesynchronised,bothinterms
ofmarketingactionsandin-houseorganisation ”.