The Big Data Challenges of Computational Market Research

April 2014

Table of Content

Toluna

What is a Market Research Study?

Toluna

Market Research Goal: Answering Questions for Brands

● ● ● ● ● ● ● ● ●

Toluna

Output Example : ‘Positioning survey’ for Hilton Garden Inn.

Toluna

Output Example : ‘Positioning survey’ for Hilton Garden Inn.

Toluna

Example : Positioning survey for Beyonce

Toluna

Example : Positioning survey for Beyonce

Toluna

Market Research Main Challenge: Targeting

●

●

Toluna

The Targeting Pipeline and Incidence Rate Targeting process

Incidence Rate: IR = Completes / Starts

Fixed set of attributes known beforehand

Free style attributes, usually unknown.

Cost is a growing function of IR

Start

Demographics

Behavioral

Study

Complete Select the right population based on simple demographic attributes: Age, Gender, Region, Ethnicity, Income, etc.

Further select based on behavioral and custom attributes: fly more than 5 times a year, uses aspirin on a daily basis, etc.

Toluna

Why is targeting hard?

Smokers

● ● ●

Tried to stop London

●

Adults

Incidence rate: 0.18 * 0.4 *.15 * .3 = 0.3 % Sample size: 333,333

UK

Toluna

State of the Art: Use Known Demographic Features ● ○ Smokers

● ●

Tried to stop

Incidence rate: 1 * 0.18 * 0.4 = 9 %

Adults in the London Area

Sample size: 11,000

Toluna

New Direction: Use Known Features and Predict Unknown Features Adults in the London Area who are predicted to be smokers

● ○ ●

Smokers

Tried to stop

●

Incidence rate: 1 * 0.85 * 0.4 = 34 % Sample size: 2,900

Toluna

How to Predict Features? The Space Model Sparse Matrix containing all the attributes (integer answers to questions) we have ever asked. Demographic attributes

Features Sex, Age, Region, etc. Users

Behavioral attributes

Shirt color

Smokes?

Red Blue

Yes No

User 1 User 2 User 3 User 4 10^^9 users

10^^7 features

Toluna

The Learning Task - The Model Try to predict answer to the “Smokes?” attribute based on other attributes. Smokes?

Dog owner?

Jogger?

Overweight?

Toluna

The Learning Task - Collaborative Filtering User correlation or Feature correlation

● ● ● ● ● ● ● Toluna

First Experiments with Multiple Imputation

Toluna

First Experiments with Multiple Imputation Some Results

● ● ●

Toluna

Current Work

● ● ● Toluna

Future Work

●

● ● ●

Toluna

Q&A

Toluna