With So Much Observed Data, Why Bother Asking Questions?

One could argue that market research arose from and centers around asking questions.  How can we better serve customers?  How is our target audience responding to our latest ad campaign?  In which ways can our business expand to see maximum growth?  Questions like these (and countless others) are fundamental to what we do.  That said, many of us in the industry are increasingly questioned about why we need to ask people questions at all when there is so much preexisting data at our fingertips.  The sheer volume of data and relative ease of access has improved in recent years, and will continue to change the way we do research.  We’ve come so far so fast that it’s worth taking a step back to consider the types of data available to researchers.  In the interest of making what could be a tome into something readable in between meetings, I’d like to focus on the relationship between “observational” and “engaged” data and where this data comes from.

Observe or Engage?

Observational data is collected without actively engaging with anyone, thus making “engaged data” (a new term I’m coining, perhaps), its foil.  And observational data takes many forms, much of it collected digitally: web and mobile activity, purchase and loyalty history, social media chatter, smart TV viewership, market research panelist profiles, CRM records; the list could go on.  One of the reasons we’ve seen these types of data become frequent topics of conversation and an increasing part of research approaches is that there is just so much of it.

What has really brought it to the fore is the technology used to process, analyze, and ultimately make sense of the data.  (I resisted the urge to mention automation, machine learning, and artificial intelligence here but darn, I just did it.)  Additionally, observation has become a substitute for engagement because it has become increasingly challenging to reach the right people to ask questions of.  If you still write those grid-heavy 40-minute surveys that no one, seriously no one, wants to take, have a look in the mirror and ask yourself why this might be.

Engaged data, that which comes from actually asking questions of people, fills something of a different but also complementary role.  Perhaps first and foremost, it helps answer the “why.”  While we might typically evoke the age-old notion of quant answering the “what” and qual answering the why, I think in this case we can look at engaged data – both quant and qual – as allowing researchers to get to the “why.”

For example, understanding motivations behind purchase behaviors or loyalty to a brand are difficult if not impossible to determine from simply looking at observed data.  The flip side of the coin in this specific example, however, is that some purchase behaviors are hard to ask people about.  Think of the last time you bought toothpaste three months ago.  I bet you don’t remember much, if anything, about it.  Marrying observed data from hard to recall points with engaged data: therein lies a sweet spot.  In some cases, observed data will be more accurate and in others, engaged data.  One of the keys is knowing what you are trying to understand and choosing the best approach, often a combinative one, to get there.

Who’s on First? What’s on Second? “I Don’t Know” is on Third.

While the approach is key, the source and associated quality of data, is even more important.  Bad data in = bad insights out.  As the market research ecosystem increasingly integrates with the broader digital ecosystem, many lose sight of – or don’t even understand to begin with – where the data they are using comes from.  I’m not going to sugar coat it; that’s a real problem.  Knowing whether your data is first, second, or third-party is crucial to understand.  So, let’s take a moment to define these terms.

First party data is collected directly from people, which could be customers through CRM or loyalty tracking or members of a market research panel.  Second-party data is someone else’s first party data, obtained through a relationship and perhaps most commonly garnered with a data match or append.  Still with me?  The first two are easy but third-party data is where it gets interesting/complicated/tricky.  Third party data is aggregated from various sources, some of which are observed, some engaged, some modeled, some seemingly conjured from unknown depths, and all of which is typically acquired from a data exchange.  I was slightly snide there to make a point.  To be clear, the point is not that third-party data is inherently bad.  It is not.  Some of it may be, however.  The point is when you are dealing with third party data, especially for research purposes, you should ask questions.  Ask questions about sourcing.  Ask questions about accuracy.

 

 

Much of the third party data available on exchanges is used for marketing and activation – for serving ads and marketing to people who fit within target audiences or segments.  Using data for these purposes has, most would argue, significantly improved ad targeting and effectiveness, though it is still an imperfect science.

If I told you that I didn’t need to ask age or gender in a survey and I could guess (read: infer) based on other things I think I know about survey participants and be right about 70% or even 80% of the time, would you be impressed?  What if I could reasonably get income level brackets right about half the time, would that be cool with you?  I bet you wouldn’t be impressed.  The demographic example is a simple one, but could be extended to many other types of modeled or inferred third-party data.  In the demographic example, I’d bet those of us who engage with market research panelists are more accurate than the inferential magicians.  In other cases and with other data points, this may not be so.

What is important is to recognize and grapple with differences when we find them, such as those between observed and engaged data as well as between first and third-party data.  How can first-party data help to calibrate and improve third-party data, including for marketing and advertising purposes?  Which types of third party data are fit for which research purposes?  Which questions does it make sense to ask – and which to observe?  How can observed and engaged data be brought together effectively to best address our clients’ objectives?  I leave you, quite intentionally, with open questions, because asking questions is one of the things we do best.