Wherecamp Navigation Conference 2015 - Geo-behavioral personas for next generation marketing and beyond
1. 1 Alexei Poliakov alexei@locomizer.com @poliakov
Rocket Space, SF
B2B, London
Minority Report 2.0: Geo-
Behavioral Personas for
Next Generation Marketing
and Beyond
7. 7
MAP-BASED PARADOX
“Rather than reproducing pictures in the bran, research results
indicate that what we perceive is a systematically altered
version of reality. Part of what we “see” are the opportunities for
and costs of acting on the environment.”
THE LESS RELEVANT THE ENVIRONMENT THE SMALLER
THE DEVIATION OF THE PERCEIVED FROM MAP-BASED
REALITY…
WHICH MEANS
… “UNDERSTANDING OF PEOPLE BY ASSUMING THEY
ARE UNMOTIVATED, NON-ENGAGED AND EMOTIONLESS
14. 14
WE PROFILE PEOPLE BASED ON SCIENTIFIC
DISCOVERY
Research on spatial behavior
in live systems
Biological Intelligence Technology –
Geo-Behavioral Interest Profiling
Cell movements and interactions People movements and interactions
15. 15
RESULTING IN GEO-BEHAVIORAL INTEREST GRAPH
(GLOBAL DATABASES OF USER AND PLACE INTEREST PROFILES)
Arts
Shopping
Eating
Sports
Crafts
Office
Financial
Leisure
Auto
Travel
Transport
36
75
80
59
10
26
17
62
48
12
54
Affinity ScoreCategory
Place ProfilesUser Profiles
WEEKDAY, 4PM
Locomizer
Algorithm
ID+ lat/lon +
timestamp
ID + lat/lon +
timestamp
ID + lat/lon +
timestamp
17. 17
DIGITISED PROFILES ARE READY FOR TARGETING,
MATCHING AND MODELLING
Cinemagoer Eating lover 1 Eating lover 2
18. 0
1
2
3
4
5
6
7
8
9
10
**** *** ** *
Populationsizewiththepositiveinterest(%)
Overal Confidence Level
Gay
0
10
20
30
40
50
60
70
**** *** ** *
Populationsizewiththepositiveinterest(%)
Overall confidence level
Shopping
0
10
20
30
40
50
60
**** *** ** *
Populationsizewiththepositiveinterest(%)
Overall confidence level
Nightlife
0
10
20
30
40
50
60
70
**** *** ** *
Populationsizewiththepositiveinterest(%)
Overall confidence level
High Street fashion
shopping
SIZE OF LOCOMIZER’S AUDIENCES BY
INTEREST BROADNESS AND RANGE
Broad
Interest
Niche
Interest
Short Middle Broad
Range
Short Middle Broad
Range
Short Middle Broad
Range
Short Middle Broad
Range
‘Calvin Klein’ audience
19. Broad
Interest
Niche
Interest
DISTRIBUTION OF THE INTERESTS SCORES FOR
FOR THE MIDDLE-RANGE INTEREST
Shopping High Street fashion shopping
Nightlife
60% of the population 52% of the population
18% of the population 4% of the population
‘Calvin Klein’ audience
20. 20
PLACE PROFILING
USE CASE
Translate individual location
history (from locomizer’s data
pool) into targetable interest
profiles
Pinpoint customers with
Interests or Intents that make
them receptive to after-work
drinks targeting (based on
target persona description)
Build heatmaps based on user
profiles of people with high
affinity to after-work drinks
Discover optimal sites to target after-
work drinks crowd (18-39 yr old
professionals) by day part
How it works:
21. 21
DELIVERABLE
Pinpoint places as granular as
a street level with people
whose Interests make them
receptive to after-work drinks
targeting by hour, day, week or
month
Intelligently decide WHEN and
WHERE to run your targeting
campaign to achieve the
maximum effect
Know daily whereabouts of crowd with
high interest to after-work drinks
all day
FRI
SAT
5-9pm
MON
pm
WED
am
Heatmap will allow to:
22. 22
DEMO
The proposed interactive heatmap will pinpoint sites
(500x500m polygons with a street level granularity) with
different levels of affinity to ‘after-work drinks’ targeting by day
part
This will enable the brand to:
– Pinpoint places with people whose interests make them more receptive
to your OOH or mobile targeting campaigns
– Intelligently decide when and where to run your OHH and/or mobile
campaign
– Influence your creative recommendations, making your product more
relevant to location
23. 23
[Sample] All day (8am-11pm) heatmap shows areas
with different levels of affinity to eating/drinking
27. 27 Photo credit: by Eva Rinaldi Celebrity and Live Music Photographer
Expected Impact
Discover non-obvious sites for
targeting
Increase foot traffic to key venues
driven by campaign relevancy
Create brand uplift by selecting
optimal target sites
MAKE EVERYBODY HAPPY
33. 33
GEO-BEHAVIORAL PROFILING IN
LONDON, September-October 2014
https://demo.locomizer.com/map/London
Area: 25 km radius around London
Unique users in the sample database: 206164
Number of historic location signals: 3261665
Source: geo-tagged tweets
Statistics:
>£0.2 mln active users within M25, which represents ~2.5% of the total population in London.
Gender: 33% male, 25% female, 10% unisex, 32% not specified
Device type: 48% iphone, 19% android, 18% sent through a web browser (could be any device), 2% ipad, 6% sent
through instagram
33
34. 34
PROFILING OF PLACES BASED ON
HISTORIC AUDIENCE INTERTESTS TO
‘NIGHTLIFE’ AND ‘GAY’ ACTIVITIES
PARIS, October-November 2014
https://demo.locomizer.com/map/Paris
Area: 8 km radius around Paris
Unique users in the sample database: 59 998
Historic location signals: 1 342 370
Identified males in the sample database: 17 922
Historic location signals by identified males: 254 460
Source: geo-tagged tweets
(c) Locomizer.com April 2015 alexei@locomizer.com
35. 35
GEO-BEHAVIORAL PROFILING IN
Tokyo, July-December 2014
https://demo.locomizer.com/map/Tokyo
Area: 25 km radius around Tokyo
Unique users in the sample database: 126 752
Number of historic location signals: 479429
Source: mobile operator
35
36. 36
MCDONALD’S CASE
WHEN & WHERE to target?
WEEKDAY, 4PM-5PM
Locomizer drove both CTR and conversion rates by 50% and 30% correspondingly,
resulting in an incremental increase in footfall of 7,000 customers in MacDonald’s
restaurants in one month
Locomizer API
Locomizer partner’s
Hyperlocal Ad
Platform
place context
NEARBY McD
50%
CTR
McDonald’s made data-driven
decisions of WHEN & WHERE
to send mobile ads based on
Locomizer’s extrapolated view
of footfall by fastfood interest
and time, resulting in 50% lift
in CTRs in comparison to
non-targeted ads.
TARGET AUDIENCE
37. GEO-BEHAVIORAL MAPS CHANGE OVER TIME
(EXAMPLE: LUXURY INTEREST)
week1 week2 week3 week4 week5 week6
average
37