Why is audience measurement difficult in mobile apps?

Anindya Datta, CEOMobilewalla

Mobilewalla

• Seattle-based, venture funded company founded and run by leaders in the big-data and advertising technology space

• Pioneering audience measurement in mobile apps

• Applying ground breaking data science techniques to the largest volumetric database of mobile app market data

Traditional audience measurement relies on panels!

Extrapolation to wider audience!

Panel limitations well-known

Not enough people on the panels

Panels not harmonised

Doesn’t reflect all populations & lifestyles Audience measurement panel size

2500 (0.06%!)

Ireland – Total TV Universe4,205,000

*Source: Nielsen Television Audience Measurement 2012 RTE Television media Sales - Oct 2012

Instrument “Gateway Devices” (Cable Box, Browser) to record user behavior

Had to throw in my favorite Extrapolation cartoon

Source: xkdc.com

Panels & Popularity Persistence

Fundamental to panel driven measurement Idea of popularity persistence

Large pool of options

“small” set of popular choices

99 – 1 rule

Objects popular today popular 30-60-90 days from today

• Panel can be assumed to eventually gravitate towards the persistent popular set

App Popularities do not Persist

20 17 13

90 day Churn35%

30 day Churn15%

Top 20 US TVprograms

Top 20 US TV Progams over 90 days

90 day Churn85%

30 day Churn45%

Top 100Apps

Top 100 Apps over 90 days

55 15100App Popularities are highly transient!

Panels don’t work for Apps

Source: Programs - Nielsen Television (TV) Ratings for Network Primetime Series Apps - mobilewalla

30 day 90 day 0 day 30 day 90 day 0 day

Majority of churn due to one time special broadcasts like the Presidential Debates, Oscar’s, Grammy’s & Sporting

Events

Regular broadcast TV shows hardly demonstrate any

popularity churn

Volatility – Ranks!

8-Feb

9-Feb

10-Feb

11-Feb

12-Feb

13-Feb

14-Feb

15-Feb

16-Feb

17-Feb

18-Feb

19-Feb

20-Feb

21-Feb

22-Feb

23-Feb

24-Feb

25-Feb

26-Feb

27-Feb

0

50

100

150

200

250

300

350

400Link The Gugl (Games) MailBox (Overall)Draw Something Free (Overall) Go To Meeting (Business)

18-24 Jun 2012 16-22 July 2012 13-19 Aug 2012 17-23 Sep 2012 15-22 Oct 2012 12-18 Nov 2012 17-23 Dec 20120

50

100

150

200

250

300

350

400

The Big Bang Theory 60 Minutes Two and a Half Men NCIS*

TV Programs

Apps

Source: Programs - Nielsen Television (TV) Ratings for Network Primetime Series Apps – mobilewalla; *Note: NCIS was not aired in the week of 15-22 Oct

Mobilewalla Pioneers Audience Measurement for Mobile Apps

Rule Based Refinement

Reference Based Estimation

Audience Based Clustering

3 Step Approach

Audience data delivered in two ways

Given an App Output indexed audience demographics (like Quantcast)

Given a target demographic Output apps that provide reach into that demographic

Rapid Adoption in the Mobile Ad Industry Use Cases

● Campaign targeting

● Supply enrichment in RTBs

● Publisher Prospecting

• Ability to cross-reference audience with popularity

• Provide data in real-time

Ground breaking big data techniques

In Summary

• One of the greatest impediments towards the widespread adoption of advertising in mobile has been the unavailability of reliable audience data for apps

• It turns out that audience measurement in mobile is difficult– The inapplicability of traditional panel based measurement techniques

is a major reason

• Mobilewalla has invented techniques, based on ground-breaking big-data science, to reliably estimate app audiences and is powering targeted campaigns as well as supply enrichment at major ad technology companies