Big Data: The Big Opportunity
A new IPA report has been produced to help the audience measurement community use 'big data' more effectively.
The report also investigates the implications of big data use for agencies, advertisers, media owners and industry media currencies - and draws up a list of ten key questions that shold always be asked (see below).
'The Big Opportunity' report, written by Newsline columnist and director of Research the Media, Richard Marks, makes clear that in the context of media research and audience measurement, big data constitutes a 'revolution' in terms of the accountability and the skill-sets demanded of the people producing it.
Some supporters of big data argue that the 'what' is all that is needed and that statistically analysing the 'what' will indicate the 'why' - motivations - through correlations and pattern recognition. However, this approach can run the risk of seeing correlations or patterns that aren't there, argues the IPA report.
With this criticism in mind, The Big Opportunity is built upon a number of key questions considered vital to overcoming the concerns voiced.
According to the report, the industry should be focusing on aspects including which areas of consumer behaviour and competitor performance are covered and which are not, which universe the data represents, how much work will be needed to turn data into insight, how frequently data is made available and whether there are gaps in audience measurement.
Already, big data provides "highly granular, passive measurement of behaviour at a low cost and speed" and allows for a "far more detailed measurement of long tail digital consumption," says the report. Yet makes it clear that the market research industry is going to have change.
"Greater in-house research and analytical skills will be needed to guide the selection and interpretation of the data sets available...[and budgets must be] allocated to the most relevant and reliable datasets."
Meanwhile the balance of research agency expertise will adjust to incorporate a higher degree of "data curation" as opposed to just "data creation".
In order to maximise the opportunities presented by both big data and industry audience services, the skilled data purchaser and user - the 'data shopper' - needs to be asking the right questions.
Ten key questions to ask about big data
- What aspects of a consumer's behaviour are covered and which are not?
- What aspects of competitor's performance are covered and which are not?
- Is the methodology transparent? What Quality Control steps are in place to ensure the accuracy and integrity of the data? Who has appraised the data set?
- What universe does the data represent?
- If it is being projected to a wider universe, what methodology is being used?
- Is the data a census, derived from a balanced sample or an opt-in sample? If the latter, what was the response rate?
- Is the data longitudinal - can we examine the same people across time?
- What demographics does the data include? How are they obtained? Are they at an individual or household level?
- Is the data compliant with privacy laws? Is its availability likely to be impacted by changes in privacy laws?
- How much work will be needed to turn it from data to insight?
Ten key questions to ask about Audience Research:
- How is the sample recruited? What steps are in place to maximize response rates and sample balance?
- Is the data collected passively or using claimed recall? Both can be appropriate but have implications for granularity.
- Is the data collected from an ongoing panel or from separate samples?
- Is the data date-specific or averaged across a time period? How is audience exposure defined? GRPs, TVRs, Daily and weekly reach and frequency need to be clearly understood as they have implications for use and interpretation
- How frequently is the data made available?
- At what level of granularity is the data available?
- What analysis package will be needed to use the data? Is the data available in raw or aggregate form for in-house systems?
- What are the analysis limitations imposed by the sample size? At what level of analysis does the data become unreliable? Is this clearly flagged by the analysis package?
- Is the service subject to audit or Technical Committee oversight?
- Are there gaps in the measurement – new technology not yet covered? What plans are in place to fill these gaps?
A full copy of the report is available on the IPA's website.