Tuesday, May 14, 2013

Live from FOCI 2013: Big Data, Little Data, New Data

Presented by Larry Friedman - TNS

This engaging session provided a close look at how the NEW market research differs from market research based on the old models.  Larry provided great examples of the new technology backed up with numbers that definitely prove the point:  the new market research is more precise and more effective.  

Old Research
Collect Data
Sample Design
Questionnaire Design
Barriers and X-Tabs
Descriptive Analysis

New research is about trying to answer questions using data that already exists.   This requires a different mindset and different skillsets.

New Research
Find Data
Make Connections
Programming & Statistics
Multi-source modeling & prediction
Data Science

New Realities 
  • Old approaches are no longer the right approaches
  • Dealing with different kinds of data – big data platform
  • Data integration  & synthesizing is key
  • Triangulate data by connecting information from different data sources
  • Insight from big data may not be exact  (as researchers were trained they should be)
  • Storytelling may be a focus

New Acknowledgments
Multi-source data for exploring, interrogating, and predicting
Soft integration vs. hard integration
Exploring / interrogating vs. prediction

·      Querying multiple data sources to address specific questions or bring the story together.
·      Using multi-source data on the same platform for prediction modeling, running of “what if” scenarios, etc.
·      May use single-source data or rely on look-alike modeling. 

New Uses

Passive Listening – Using technology, we know more than ever about customers’ experiences without asking:
·      Mobile usage (apps, feature, etc.)
·      Web search and sites
·      Ad exposure
·      Location awareness
·      Audio sampling

The issue is balancing information gathering with privacy.  This is where opt-in approaches enter the consumer intelligence picture.

·      Rely less on questioning and more on listening.
·      Measurement takes on a new more targeted role.
·      Cross-platform use has exploded, but mobile technology lets market researchers track these behaviors.

Thin tracking is the result of greater access over longer periods of time.  Think tracking, plus social media data, plus click stream data, plus marketing inputs, plus…and so on.

In-market, on the fly testing because we can take pro-active action with consumers who often become partners in the process. 

A lot of guessing makes up the survey and questionnaire process.  New research goes beyond this recognition and memory based work.  Move beyond memory laden survey metrics.  Nielsen TV tracking would be an example. 

1.     Recognition
2.     Opportunity to see
3.     Digital media effectiveness

Tag digital ads and see if actual exposure impacts the business.  Track online exposure and tease out those consumers that are known to have seen an ad.

Observation of tagged digital ads can provide really explicit information.  For example: Exposure to ads between 7 and 9 were not effective.  The technology lets market researchers track what people are exposed to (for example, using watermarks on ads) and then receive a ping when an exposure to an ad has occurred.   

Key take-aways:  Traditional recognition metrics can be misleading.  
·      People would say that they had seen an ad, when, in fact, they had not.
·      The research tends to overstate the effect when looking at memory-laden research.
·      The attitudinal measures can be coupled with the actual behavior of consumers as a result of this new exposure tracking.  Studies are showing brand lift from ad exposure – now that it can be tracked with accuracy. 
Digital Segment Targeting – delivers more impressions to the right consumers.

Look-alike models allow brand growth target via digital media.  Example, 20% of the ads can result in 60% of the customers.

Come one has to do something…It’s just incredibly pathetic that has to be us.” 
~Jerry Garcia
Gigi DeVault writes a market research column for About.com

No comments: