2. THIS REPORT FOCUSES ON HELPING
MARKETERS UNDERSTAND:
The key drivers to
location accuracy
The current state of
location data quality in
the mobile ecosystem
The opportunities and
obstacles of mobile
location data
MARKETER TOOLS ADAPT WHILE
UNDERLYING DATA REMAINS A CHALLENGE
METHODOLOGY
This report marks the fourth edition of Thinknear’s Location Score Index
(LSI). When we first published the Location Score Index in Q2 2014,
our goal was to bring transparency to the mobile industry in an effort to
encourage industry-wide improvements in data quality. While much has
changed in the last year, location data accuracy still remains an issue
for marketers.
Our most recent findings indicate that while the overall programmatic
space continues to grow very quickly, the quality of location data remains
relatively unchanged from a year ago, on average. As predicted, mobile
publishers are flocking to programmatic and are also sharing generous
amounts of location data. Unfortunately, publisher behavior is slow to
change, and predictions that the quality of publisher location data would
improve rapidly have yet to be realized.
Thinknear, along with other ad-tech players, is currently working with the
IAB and the MMA to define industry standards and make adjustments to
the OpenRTB specifications which will ultimately improve data reliability
from publishers. Additionally, many in the industry have begun work to
develop tools similar to Thinknear’s Location Score platform that allow
marketers to filter out poor quality location data. The necessity for
this technology has forced some providers out of the space and leaves
marketers with a small handful of location-based marketing platforms
capable of handling the complexities of the current mobile ecosystem.
Looking forward, we still anticipate that data quality will improve with
time and adjustments to industry standards. We will begin publishing the
Location Score Index on a semi-annual basis, while continuing to focus
our efforts on bringing clarity to key mobile marketing issues. Our goal is
to help marketers understand both the opportunities and obstacles that
the mobile marketing world faces.
LOCATION
SCORE INDEX:
A YEAR IN
REVIEW
The Thinknear platform accesses
the largest sources of location-
enabled mobile inventory in the U.S.
To compile this report, we sampled
and analyzed data from more than
one billion ad impressions and ran
location accuracy tests on more than
500,000 consumer ad experiences.
2 Location Score Index Q1 2015 twitter @Thinknear
3. A YEAR IN CHARTS
eMarketer reported a nearly 400% increase in U.S. mobile RTB
ad spend in 2014 and has forecasted a nearly 300% increase
in 2015. The charts below further highlight growth in available
RTB inventory.
LOCATION-ENABLED INVENTORY IS ALSO INCREASING, KEEPING
PACE WITH OVERALL PROGRAMMATIC GROWTH. THE GROWTH IN
LOCATION-ENABLED INVENTORY ALLOWS CAMPAIGNS WITH VERY
SPECIFIC LOCATION TARGETING GOALS TO BE RUN AT SCALE.
QUARTERLY PROGRAMMATIC
INVENTORY
(INDEXED TO Q2 2014)
QUARTERLY
LOCATION-ENABLED
PROGRAMMATIC INVENTORY
(INDEXED TO Q2 2014)
Q3 Q4 Q1Q2
3
2.5
2
1.5
1
0.5
Q3 Q4 Q1Q2
3
2.5
2
1.5
1
0.5
PROGRAMMATIC INVENTORY CONTINUES TO INCREASE. OVER THE
LAST YEAR, THE INDUSTRY SAW A 2–3X GROWTH IN AVAILABLE
INVENTORY. AS MANY ANALYSTS PREDICTED, PUBLISHERS ARE
FLOCKING QUICKLY TO PROGRAMMATIC PLATFORMS.
PROGRAMMATIC AD BUYING HAS
EXPERIENCED TREMENDOUS GROWTH
OVER THE LAST YEAR.
62% OF AD REQUESTS INCLUDED
LOCATION DATA IN Q1 2015,
WHILE IN 2012, ONLY 10% OF
ALL AD REQUESTS CONTAINED
LOCATION DATA.
twitter @Thinknear3 Location Score Index Q1 2015
4. 50
LOCATION SCORE
™
51 (Q4)
VOLUME BY ACCURACY LEVEL (Q1, Q4):
37% HYPER LOCAL
Location data was accurate to within 100 meters
of the user’s true real-time location.
(Size of a football field)
9% LOCAL
Location data was accurate between 100 and
1,000 meters of the user’s true real-time
location. (Approx. 0.6 miles)
26% REGIONAL
Location data was accurate between 1,000 and
10,000 meters of the user’s true real-time
location. (Approx. 6 miles)
18% MULTI-REGIONAL
Location data was accurate between 10,000 and
100,000 meters of the user’s true real-time
location. (Approx. 60 miles)
10% NATIONAL
Location data was not accurate to within
100,000 meters of the user’s true real-time
location. (Greater than 60 miles)
37%
10%
28%
17%
8%
UPDATED
INDUSTRY SCORE
LOCATION SCORE BY
EXCHANGE
We score location on a 100-point, non-linear scale. Thus, it’s easier for the
industry score to grow from 25 to 35 than it would be to grow from 75 to 85.
We constantly update the algorithm used to calculate Location Score, so industry
scores fluctuate from time to time.
66
TOP QUARTILE
39
BOTTOM QUARTILE
70
60
50
40
30
0 Mobile exchanges are not equal in terms of data quality. The chart
on the left shows the average Location Score for the top and bottom
quartiles of exchanges. The score varies from the high 30s to the high
60s. Marketers need tools to identify the most accurate data within
each exchange.
4 Location Score Index Q1 2015 twitter @Thinknear
5. THE INDUSTRY LOCATION
SCORE HAS REMAINED LARGELY
UNCHANGED OVER THE COURSE
OF THE PAST 12 MONTHS.
INTERPRETING THE RESULTS
Over the last year, the average industry-wide
Location Score has ranged from 49 to 55,
indicating that publishers are still having difficulty
sending high-quality location data through mobile
programmatic exchanges.
Growth in the overall programmatic space means
marketers have plenty of high-quality inventory
available, but finding it requires location scoring
technologies to filter out all but the most reliable
data sets.
In the last two quarters, our analysis indicates
that a large portion of lower-quality location
data is attributable to apps that are new to
the programmatic ecosystem. We also noted a
marked increase in the volume of mobile web
traffic entering the programmatic space (as
opposed to mobile app traffic). Mobile web has
traditionally had lower quality location data, which
appears to be impacting the industry as well.
INDUSTRY LOCATION SCORE TREND
WHAT DOES THIS MEAN FOR MARKETERS?
As the mobile world becomes more complicated, mobile marketers should be aware of the existence
and impacts of high-quality and low-quality location data. When planning and executing location-based
campaigns, mobile marketers need to ask the right questions:
INDUSTRY LOCATION SCORE TREND
What is the source
of the location data?
If the data is scored
or filtered, how is it
being done?
Is the data taken at
face-value or is it
scored for accuracy?
Q3 2014 Q4 2014 Q1 2015Q2 2014
60
50
40
55 51 5049
5 Location Score Index Q1 2015 twitter @Thinknear
6. As this report demonstrates, app and mobile web publishers often send inaccurate location data
within their ad requests. There are multiple reasons why this happens and we’re frequently asked
if the issue is driven by publisher fraud. While there will always be some element of intentional
fraud when there are large sums of money involved, our analysis and interactions with inventory
suppliers indicate that most of the inaccurate location data in mobile is driven by a lack of
standards and a complicated ecosystem that most publishers aren’t able to effectively navigate.
WHY DO PUBLISHERS STRUGGLE
WITH LOCATION DATA?
LOCATION INACCURACY:
IS IT FRAUD?
1. LACK OF KNOWLEDGE WORKING
WITH SUPPLY-SIDE PLATFORM SDKS:
Publishers are focused on building their applications; advertising
technologies often become a secondary priority. Publishers
have significant control over the data they pass as part of an ad
request, but linking specific data to relevant fields in the RTB
specification requires diligence on the part of the publisher.
2. THE ASSUMPTION THAT ANY
LOCATION DATA IS BETTER THAN
NONE AT ALL:
Publishers assume that marketers want any kind of location
data, even if it is stale (pulled hours or weeks in the past) or not
representative of a user’s current real-time location. However,
publishers don’t realize that sending inaccurate location data
impacts the user experience and results in fewer advertisers
willing to purchase the inventory.
3. UX DRIVEN DATA DECISIONS:
Most publishers are genuinely interested in supplying
accurate data, but do not want to negatively impact the user
experience. Impacts of pulling high-quality location data can
include introducing latency, utilizing too much network traffic,
significantly draining the battery, or popping up a location
request prompt. As a result, publishers sometimes opt for
location data with subpar quality over a negative impact to
user experience.
4. MALICIOUS INTENT:
Publishers know that attaching geo-data to an ad impression will
have a positive impact on eCPMs. While rare, some publishers
may intentionally provide false location data in an effort to
generate higher revenues.
6 Location Score Index Q1 2015 twitter @Thinknear
7. EDUCATION THROUGHOUT
THE VALUE CHAIN:
Demand-side buyers need to emphasize
the value of quality location data to
their supply-side partners. Supply-side
platforms that represent the inventory
of publishers in turn need to educate all
of their publishers.
USER REGISTRATION
AND CENTROID DATA:
User registration data often captures
a home zip or DMA. These locations
are often converted to “centroids”
and passed in the ad request as the
center point of the provided zip code.
While relevant in a broad context,
these data do not provide any reliable
context for behavioral or real-time
marketing strategies.
AS DEMONSTRATED BY THE LOCATION SCORE INDEX, THE INDUSTRY HAS A LONG WAY TO GO IN
IMPROVING DATA ACCURACY. BELOW ARE SOME OF THE STEPS THAT WOULD HELP IMPROVE THE STATE
OF THE INDUSTRY:
STRICT SUPPLY-SIDE CONTROL
OVER LOCATION DATA:
Supply-side platforms often leave it to
publishers to pull location themselves
and include the data in outbound ad
calls—rather than the SSP’s SDK to
fetch location data. More consistent
use of SDKs with advanced location
tools will help improve accuracy.
IP-BASED
LOCATION DATA:
IP data is typically based on the server
location of an app and has nothing to
do with the user’s true location. It is
often used because it requires no user
permissions and it can be done on the
server side, thus requiring no extra
client-side work. However, IP-based data
is a poor tool for mobile targeting given
its lack of relevance to the mobile user’s
actual location.
CACHED DATA:
Some publishers only fetch location
data periodically in an effort to
conserve battery life. The duration
of time between location “pulls” can
range significantly. Marketers seeking to
leverage data about a user’s real-time
location or the locations a user has
been in the past need the freshest data
possible. Many industry trade groups
are exploring changes to the OpenRTB
spec that would require publishers to
communicate the “freshness” of data.
INACCURATE DATA CAN COME IN MANY
FORMS. BELOW ARE A FEW EXAMPLES OF HOW
INACCURATE LOCATION DATA CAN BE PRESENTED:
IMPROVED DETAIL AND
TRANSPARENCY IN SPECS:
The OpenRTB specification should be
updated to include references to when
a location was fetched if the device’s
Location Services were used.
WHAT DOES INACCURATE DATA
LOOK LIKE?
7 Location Score Index Q1 2015 twitter @Thinknear