One of the crucial trends nowadays will be the growth of attribution modeling. However, many savvy marketers are still missing the opportunity to reap the benefits of attribution modeling, hence this whitepaper is aimed at providing a comprehensive overview of what marketing attribution is all about.
2. Foreword
If we are to summarize the current top 10 marketing topics around the world, attribution modeling must be one of
the latest hot topics within the industry. According to the report conducted by Econsultancy and Adobe in November
2012, only 26% of advertisers worldwide and 36% of advertising agencies carried out marketing attribution. It is truly
evident that attribution modeling is rapidly growing and its value is being recognized by a vast majority of marketers
nowadays. Based on the report, 29% of advertisers that do measure attribution say it has significantly benefited their
business while 60% of those realized that it is beneficial. Marketers can better understand their budget allocation,
gain insights into affiliate marketing effectiveness and optimize their marketing mixes.
However, attribution modeling is still at its initial stage of gaining popularity within China. Many of the marketers are
not familiar with attribution modeling and the purpose of this whitepaper is to elaborate on the use of attribution
modeling and demonstrate its effectiveness on measuring the success of your marketing efforts.
Co-founder of iClick – Ricky Ng
1
3. Table of Contents
Part One
Why do marketers need attribution modeling?
- Why do marketers need attribution modeling?
- What is the biggest value of attribution modeling?
Part Two
What is attribution modeling?
- What is attribution modeling?
- Commonly used attribution models
Part Three
How attribution modeling creates value to marketers?
- Case analysis
- How to make use of attribution modeling to analyze your campaign?
Part Four
An introduction to XMO attribution modeling
2
4. Part One
Why do marketers need attribution modeling?
For many marketers, managing marketing budget and maximizing ROI (Return On Investment) are the biggest
objectives of every campaign. With the challenges faced by marketers on the complexity of media channels, it is
relatively difficult to conclude the ideal channel to be used. When you plan to utilize your marketing budget for a
campaign, will you have the following difficulties?
- How to optimize channel mix to maximize ROI?
- Which channel is most suitable for my products?
- How do you manage the whole media buying process to measure the effectiveness in real time?
- Will there be any impact to the channel in the case of adding or reducing marketing budget?
- Any difference on user profile from different media channels?
- Does this channel truly bring my marketing campaign with the most desirable results?
Why does attribution modeling gradually become critical to marketers? Below is an example to help you easily
understand:
For example, assume 10 successful conversions are received in a week from search engine marketing and display
advertising. If marketers could only track the conversions based on last click, then the result would show:
Last Click Conversion
1
Search Engine
2
Search Engine
3
Search Engine
4
Search Engine
5
Display Ad
6
Display Ad
7
Search Engine
8
Search Engine
9
Search Engine
10
Search Engine
As the above table illustrates, marketers may think that the conversion rate appears to be extremely low for display
advertising. In this case, marketers will intuitively allocate more budget to search engine marketing.
3
5. However, with end-to-end conversion tracking available, marketers can now better understand the process behind
each conversion more thoroughly:
Last Click Conversion
Conversion Path
1
Search Engine
Display Ad → Search Engine
2
Search Engine
Display Ad → Search Engine
3
Search Engine
Display Ad → Search Engine
4
Search Engine
Search Engine
5
Display Ad
Display Ad
6
Display Ad
Display Ad
7
Search Engine
Display Ad → Search Engine
8
Search Engine
Display Ad → Search Engine
9
Search Engine
Display Ad → Search Engine
10
Search Engine
Search Engine
Based on the analysis of conversion path, online display advertising has actually assisted 6 out of the 8 conversions
made from search engine marketing.
Do marketers now have a new perception of online display advertising based on its results? By learning the insights
from the conversion path report, marketers can now understand more about the “assist effect” amongst different
advertising channels.
The main advantage of attribution modeling is to reflect the real value of the contribution made to each channel. For
example, if 1 credit will be given only if the channel has contributed in a conversion, in this case, the value of search
engine marketing and display advertising will be calculated as follows:
Last Click Conversion
Conversion Path
Search Engine
Display Ad
1
Search Engine
Display Ad → Search Engine
1
1
2
Search Engine
Display Ad → Search Engine
1
1
3
Search Engine
Display Ad → Search Engine
1
1
4
Search Engine
Search Engine
1
0
5
Display Ad
Display Ad
0
1
6
Display Ad
Display Ad
0
1
7
Search Engine
Display Ad → Search Engine
1
1
8
Search Engine
Display Ad → Search Engine
1
1
9
Search Engine
Display Ad → Search Engine
1
1
10
Search Engine
Search Engine
1
0
8
8
Total
4
6. The biggest value of Attribution Modeling
The above example demonstrates the importance of attribution modeling. Not only it helps the marketers with the
ability to monitor performance of different media channels, but it can also increase ROI (Return On Investment)
significantly by re-allocating budget amongst channels. With some critical points mentioned above, more and more
marketers are willing to adopt attribution modeling in the long run.
44
%
54
%
%
58
63
%
72
%
Attribution Modeling - Benefits to Marketers
Able to reallocate
budget across
channels and
improve ROI
Better understand
how digital channels
work collaboratively
Gained valuable
insights into
consumer makeup
and behavior
To provide objective
reasons to justify
more marketing
budget
Data Source: Google analysis <Marketing Attribution: Valuing the Customer Journey>
5
Increase understanding
on the role of
online/offline media
interactions
7. Part Two
What is Attribution Modelling?
Attribution modeling is a set of rules that determines the sales or conversions assigned to each touch point in
conversion paths.
- Savvy marketers understand that you don’t attract your customers with just one simple message, just one image,
or just one ideally positioned advertisement. It is a complicated process of planting the seed, nurturing it and at
last harvesting the fruits of your marketing efforts.
- Right before your customers buy or convert, they may encounter various parts of your online marketing campaign
including paid and organic search, email, affiliate marketing, display ads, mobile placements, etc. Each of these
factors has a significant impact on the campaign results.
- In the past, the value of the marketing campaign is only determined by the last click of the users in the media
channel. But when you think thoroughly, the user may be affected not just on his/her last click but also in other
areas during the conversion. Attribution Modeling offers a series of different methods to determine the role that
channels play throughout the conversion journey.
Day 1: User views ad on website............................................................................
Cause
Day 2: User search for “smartphone”......................................................................................................
Promote
Day 4: User clicks on Banner Ad....................................................................
Promote
Day 35: User sees a video ad of a smartphone brand..........................................................................
Promote
20 mins later: User searches for that smartphone brand...........................
conversion
Before the user completes a conversion, they will usually encounter with various advertisements many times.
Some media channels help to expose the ad to the users, while others may use to assist conversion.
6
8. Commonly used attribution models
To optimize the use of attribution modeling, marketers must first clearly understand and identify its marketing
objectives regardless of whether to increase conversion quantity, increase user registration or download discounted
coupons, etc.
In the meantime, marketers should also understand how credit should best be assigned to each conversion funnel.
Below are the most commonly used attribution models and its appropriate use:
Last-Touch Attribution
0%
The Last-Touch Attribution Model attributes 100% of the conversion
10
value to the last referring channel with which the customer comes
across before purchasing or converting. This model is extremely
common among most marketers, hence it is a great baseline for
comparison with other models.
Appropriate use: If your advertisements and campaigns are
developed to drive customers traffic at the time of purchase, or your
business is mainly transactional with a sales cycle that does not
involve a consideration phase, the “Last-Touch Attribution” model
The direct conversion
value to the last referring
channel
PAID AD
DIRECT
conversion
value
100%
10
0%
may be ideal.
Last-Touch Model
ORGANIC REFERRAL
SEARCH
First-Touch Attribution
The First-Touch Attribution model attributes 100% of the conversion
value to the first channel with which the customer comes across.
Appropriate use: If your advertisements or campaigns are developed
to maintain awareness with the customer, this model fits you
perfectly. For example, if your brand is not commonly known, you
may primarily focus on exposing your brand to customers through
keywords or different channels.
7
First-Touch Model
ORGANIC REFERRAL
SEARCH
conversion
value
100%
PAID AD
DIRECT
Credits received from
organic search is based
on the first interaction.
9. Time-Decay
If the sales cycle involves only a short consideration phase, then this model may be appropriate for you. It will assign
the most credit to touchpoints that occurred closest to the time of conversion.
Appropriate use: If you run one-day or two-day promotional campaigns, you may wish to assign more credit to
interactions during the promotion. In this case, interactions that occurred a week before have smaller value as
compared to those touchpoints near the conversion. The “Time Decay” model may be appropriate for assigning
5%
15
%
30
%
50
%
touchpoints during the day or two to generate conversions.
Time-Decay Model
ORGANIC
REFERRAL
SEARCH
5%
DIRECT
30%
15%
PAID AD
50%
Position-based
This model allows you to use a combination of “First-touch Attribution” and “Last-touch Attribution” models. Other
than assigning all the credits to either the “First-Touch Attribution” or “Last-Touch Attribution”, you can easily allocate
the credits between them. One commonly seen scenario is to assign 40% credit each to the first and last interactions,
and assign the remaining 20% credit to the middle interaction.
Appropriate use: If you primarily value touchpoints that sell your brand to your customers and final touchpoints that
10
10
%
%
30
%
50
%
resulted in sales, you can simply use the “position-based”model.
Position-Based Model
ORGANIC
REFERRAL
SEARCH
30%
8
DIRECT
10%
10%
PAID AD
50%
10. Part Three – How attribution modeling creates value to marketers?
Case Analysis
- Customer:
Vertical e-Commerce
- Objective:
To increase online conversions
- Cost Per Acquisition (CPA):
$380
- Promotional Period
Oct to Dec 2012
- Promotional Results
See below
Channel
Consumption
Conversion
Share
CPA
Google – Paid Search
$512,236
1,980
31.9%
$259
Google – Organic Search
-
138
2.2%
-
Baidu – Paid Search
Baidu – Organic Search
$1,403,232
2
3,123
-
29
Ad Network
$308,983
Ad Exchange
Grand Total
50.3%
4
3
1
$449
0.5%
-
178
2.9%
$1,736
5
$63,459
760
12.2%
$84
6
$2,287,910
6,208
100.0%
$369
Assume the marketer has no assist from attribution modeling, the results on every channel will be indicated as
follows:
1. Google – Paid Search: CPA is ideal, but recommend to increase budget
2. Google – Organic Search: Low conversion channel generating poor performance
3. Baidu – Paid Search: Key conversion driver, but CPA exceeded and then require to lower CPA
4. Baidu – Organic Search: Low conversion channel generating poor performance
5. Ad network: High CPA but low conversion, exceeded the target CPA, hence recommended to drop this channel
6. Ad exchange: CPA is ideal but low conversion and require to increase budget
Note: The performance figures are specific to this particular case study only. It does not represent the actual performance of the advertising channel in general.
9
11. But with the data provided by attribution modeling, the conversion value of each channel will be truly reflected as
follows:
Conversion
Channel
Effective
Consumption
Conversion
Share
CPA
Conversion
Share
Real Conversion CPA
Google – Paid Search
$512,236
1,980
31.9%
$259
2,099
33.8%
Google – Organic Search
-
138
2.2%
-
287 2
4.6%
-
Baidu – Paid Search
$1,403,232
3,123
50.3%
$449
2,820 3
45.4%
$498
Baidu – Organic Search
-
29
0.5%
-
233 4
3.8%
-
Ad Network
$308,983
178
2.9%
$1,736
590 5
9.5%
$524
Ad Exchange
$63,459
760
12.2%
$84
179
2.9%
$355 6
Grand Total
$2,287,910
6,208
100.0%
$369
6,208
100.0%
$244 1
$367
Main changes of each channel:
1. Google – Paid Search: CPA remains ideal, and effective conversions increased slightly
2. Google – Organic Search: Effective conversion has increased, in other words, this channel can assist other channels
for conversion
3. Baidu – Paid Search: Effective conversion has lowered, in other words, this channel relies heavily on other channels
for conversion
4. Baidu – Organic Search: Effective conversion has increased, in other words, this channel can assist other channels
for conversion
5. Ad network: Effective conversion has increased significantly, in other words, this channel is mostly used for
assisting other channels
6. Ad exchange – CPA has increased significantly, indicating such channel belongs to the last part of the conversion
funnel
10
12. How to make use of attribution modeling to analyze your campaign?
When marketer gained a better idea of the actual conversion value from each channel through attribution modeling,
how he/she should practically apply the insights to optimize the campaign?
Attribution modeling goes beyond just the first step of optimization for marketers. In the next step, marketers should
look for relevant insights from other reports. In this case study, the marketer has the following insights and actions:
Insight 1:
Conversion
Channel
Effective
Consumption
Conversion
Share
CPA
Conversion
Share
Real Conversion CPA
Google – Paid Search
$512,236
1,980
31.9%
$259
2,099
33.8%
$244
Google – Organic Search
-
138
2.2%
-
287
4.6%
-
Baidu – Paid Search
$1,403,232
3,123
50.3%
$449
2,820
45.4%
$498
Baidu – Organic Search
-
29
0.5%
-
233
3.8%
-
Ad Network
$308,983
178
2.9%
$1,736
590
9.5%
$524
Ad Exchange
$63,459
760
12.2%
$84
179
2.9%
$355
Grand Total
$2,287,910
6,208
100.0%
$369
6,208
100.0%
$367
The actual amount of conversion has decreased in
Rank
Baidu Paid Search after analyzed by attribution
1
Baidu – Paid Search
143%
modeling, in other words, the conversion from this
2
Google – Paid Search
32%
channel has in fact benefited from the contribution by
3
Google – Paid Search → Ad Exchange
9%
other channels. According to the XMO conversion path
4
Ad Network → Baidu – Paid Search
4%
5
Baidu – Organic Search → Baidu – Paid Search
3%
6
……
report, 7% of the conversion comes from Ad Network
and Baidu Organic Search to Baidu Paid Search. Hence,
Conversion path report
Share
……
the focus of the optimization should lie on the basis of
Baidu Organic Search - Search terms report (Simplified version )
these two channels instead.
Search terms
Cost
Conversion should take place as below :
Online Shop
$ 54,935
184
3.5%
$299
Baidu Organic Search:
Online Shopping Site $ 54,618
137
2.6%
$399
Shopping Discounts
$ 30,702
74
1.4%
$415
Free Delivery
$ 18,205
58
1.1%
$314
Discounted coupons $ 10,135
47
0.9%
$216
Auction
$ 17, 691
37
0.7%
$478
Other Search Terms
$ 1,729,183
4,733
89.8% $365
Grand Total
$ 195,468
5,270
100%
Based on the keyword search report on the left,
marketers should ensure conversion driven keywords
are included in the paid search campaigns so as to keep
a constant conversion rate.
11
Conversion Share
CPA
$363
13. Ad Network:
Ad Network Conversion Performance report
Region
As Ad Network is defined as the top-notch “assisting
Cost
Conversion
Share
CPA
channel”, it is mostly capable of arousing brand
Beijing
$ 372,417
1,247
20%
$299
awareness to acquire new customers.
Shanghai
$ 336.163
1,093
18%
$308
To enhance the effectiveness of Ad Network, marketer
Guangdong $ 332,110
963
16%
$345
Chengdu
$ 312,168
749
12%
$417
Other
$ 953,052
2,156
35%
$434
Total
$ 2,287,910
6,208
100%
$369
can re-allocate their marketing budget based on
conversion district, as such to allocate more budget to
cities with higher conversion rates such as Beijing,
Shanghai, etc.
Insight 2:
Conversion
Channel
Effective
Consumption
Conversion
Share
CPA
Conversion
Share
Real Conversion CPA
Google – Paid Search
$512,236
1,980
31.9%
$259
2,099
33.8%
$244
Google – Organic Search
-
138
2.2%
-
287
4.6%
-
Baidu – Paid Search
$1,403,232
3,123
50.3%
$449
2,820
45.4%
$498
Baidu – Organic Search
-
29
0.5%
-
233
3.8%
-
Ad Network
$308,983
178
2.9%
$1,736
590
9.5%
$524
Ad Exchange
$63,459
760
12.2%
$84
179
2.9%
$355
Grand Total
$2,287,910
6,208
100.0%
$369
6,208
100.0%
$367
After the analysis of attribution modeling, the actual
Rank
conversion coming from ad exchange platform drops
1
Baidu – Paid Search
143%
from 12.2% to 2.9%, which means such channel
2
Google – Paid Search
32%
benefits from other channels in order to get the
3
Google – Paid Search → Ad Exchange
9%
conversion.
4
Ad Network → Baidu – Paid Search
4%
5
Baidu – Organic Search → Baidu – Paid Search
3%
6
……
Conversion path report
Share
……
From the conversion path report, we can derive that the Google Paid Search is mainly the assisting source for ad
exchange platform. Although Google Paid Search itself is an extremely effective channel, there are still some users
who fail to be converted through this channel alone. Hence, it is recommended to adopt search retargeting to ensure
the campaign running in ad exchange can capture the losing users from Google Paid Search.
12
14. Insight 3:
Conversion
Channel
Effective
Consumption
Conversion
Share
CPA
Conversion
Share
Real Conversion CPA
Google – Paid Search
$512,236
1,980
31.9%
$259
2,099
33.8%
$244
Google – Organic Search
-
138
2.2%
-
287
4.6%
-
Baidu – Paid Search
$1,403,232
3,123
50.3%
$449
2,820
45.4%
$498
Baidu – Organic Search
-
29
0.5%
-
233
3.8%
-
Ad Network
$308,983
178
2.9%
$1,736
590
9.5%
$524
Ad Exchange
$63,459
760
12.2%
$84
179
2.9%
$355
Grand Total
$2,287,910
6,208
100.0%
$369
6,208
100.0%
$367
From the data analysis combining attribution modeling
Rank
and conversation path, it is obvious that Google Paid
1
Baidu – Paid Search
143%
Search is a highly effectively channel, whether it serves
2
Google – Paid Search
32%
as a conversion channel itself or assists other
3
Google – Paid Search → Ad Exchange
9%
advertising channels to convert.
4
Ad Network → Baidu – Paid Search
4%
5
Baidu – Organic Search → Baidu – Paid Search
3%
6
……
Therefore marketer may consider to focus on the
optimization effort for this single channel, for example,
to perform keyword expansion to boost the traffic
volume.
13
Conversion path report
Share
……
15. Conclusion:
Attribution Modeling may not be able to answer all the questions arose from online marketing. Marketers should
make use of other in-depth data analysis and continuously try out different optimization methods in order to
constantly improve their campaign performance. Below are some suggestions for the marketers:
Identify your marketing objective
Establish a clear and solid marketing objective is the first step to success. Is your primary objective based on attracting
new users, improving conversion directly or increasing the return rate of existing clients? Different objectives
determine different attribution models to be selected.
Refer to other reports to generate practical insights
Attribution modeling does not mean automation of all optimization procedures. Marketers should review other
reports (such as conversion path, conversion keyword reports, etc.) to conclude a feasible optimization solution.
Attribution modeling can help understand their product’s conversion funnel
Apart from increasing the conversation quantity, we suggest marketers to try shortening conversion time, for
example, add redirect function, or offer discounts to those returning users generated from the ad as to increase the
chance of conversion.
Keep trying out different optimization strategies
Invest time to try out different optimization strategies and study their effectiveness instead of focusing on building a
perfect attribution modeling.
14
16. Part Four
An Introduction to XMO attribution modeling
iClick’s cross-channel optimization platform, XMO is capable to generate predictive analytics to assist marketers in
marking data-driven marketing decisions. Our product design aims at every detail from ad tracking, data analysis to
optimization. Our product edge include:
- Tracking code:
Capture every interaction of the user, from their first exposure to the ad until the last conversion completed to allow
marketer to better understand customers’ online behavior.
- Integrate conversion path with cross-channel analysis:
Provide channel assist report to analyze the relationships among different channels so as to understand the real
value of the channels.
- Provide multi-dimensional and customized conversion funnel analysis:
Marketer can customize their unique conversion funnel to cater to different features of their promotional products,
aiming at increasing the relevance of data analysis.
- Various types of attribution modeling:
Use different attribution modeling on different promotional products, allowing the data analysis to better reflect
the actual performance.
XMO user can download attribution modeling reports from a range of attribution modeling
15
17. Assist Effect Report – Better understand the real value of channel conversion
Conversion Path – Understand the
frequent conversion path can allow
channel conversion
About iClick
iClick is the leading digital buy-side platform in Asia that integrates search,
display, social media and mobile marketing capabilities to allow marketers
to plan and manage their cross-channel campaign in one single
destination.
Harnessing the power of programmatic buy and ad technology, the
proprietary platform XMO (*Cross-marketplace Optimization Platform) is
designed to deliver maximum ROI and greatest efficiency to marketers.
Visit www.i-click.asia and follow us at weibo.com/iclickasia
Sales enquiry: sales@i-click.asia
16