Download the
NEW e-book from GreenBook: Insights That Work - best research from the GRIT Top 50 companies | DOWNLOAD NOW

Cross-Platform Data: Where Sound Bites Meet Research Reality

reality-check (1)

 

By Florian Kahlert

At a recent research conference, a panelist (representing an agency that shall remain unnamed) was yet again proclaiming the death of panel data and advocating a move to cross-platform, census-level data for planning and buying.

I had to bite my tongue, as I often do in these cases. Actually, in theory, I agreed with him. If we had reliable cross-platform data for TV, radio, print, online and mobile for the same individuals, accurately linked and at levels that approach census-level (millions, not thousands) — and then could connect that data to the same person’s product ownership and consumer behavior — we would be in advertising Nirvana. (Or, Orwell’s 1984.)

The idea of “truly 360-degree measurement” is not new, and attempts to get closer to it are always appearing; but they bump up against hard, cold realities. Let’s talk about a few:

First, that same agency person who would be so delighted by the abilities of this system would likely be unwilling to pay for the services. To do what he envisioned would be prohibitively expensive.

Second, what sounds awesome as a conference sound bite (or written in a blog post, like this one) is technologically incredibly complex, massively big, and – unless you are the NSA, with virtually unlimited funds – extremely hard to do. To combine passive data for the same person reliably across multiple platforms at scale is something Richard Branson might consider beyond rocket science.

Just to provide a simple detail – managing a passive panel (not even census) requires constant oversight as people change devices, move to different states, and more. It also demands ongoing software development, as mobile companies change the way they do things; and it generates terabytes of data every day that need to be cleaned, processed, and made actionable by running them through ever-evolving taxonomies. And that is just for one platform; do this across digital and TV, and you are multiplying the complexities.

Third, we have not even talked about the definition of “census-level.” Does it mean all 200 million US adults? Is a sample of 20 million enough? In other words, do big numbers without actual sampling methodology truly “represent”? What if my 20 million represent a populous of mostly high-income people living in big metro areas? Is that “good enough”?

A path forward

Now, just being a nay-sayer is not really helpful. Let us look at some things we can actually do.

First, there are companies out there that have excellent program level TV data – Nielsen and Rentrak come to mind.

Second, other companies have awesome product ownership and print media consumption information (such as yours truly, GfK MRI); but this is mostly recall data, not passive.

Thirdly, there are companies like Tapad that do an excellent job of connecting different devices (mobile online), but they do not know much about the person’s TV viewing habits or OOH behavior.

The key to a way forward is to acknowledge that no one can afford to own and generate all the data anymore by themselves; we need to find ways to combine different data sets to come up with a more universal view. And the only way I see to do this that will not bump up against Orwellian levels of privacy intrusion is in anonymized matching, and modeling unmatched sets, and then calibrating them all against representative, carefully managed reference panels. And there again, you have the need for panels — the very opposite thing of census data.

Needless to say, we are constantly working on it. Living in the real world is no small task, it seems.

Please share...

5 responses to “Cross-Platform Data: Where Sound Bites Meet Research Reality

  1. Florian I am genuinely interested to hear your opinion to my comments below in CAPS next to the issues you outline.

    Thank you
    Michalis

    First, that same agency person who would be so delighted by the abilities of this system would likely be unwilling to pay for the services. To do what he envisioned would be prohibitively expensive. I CHALLENGE EVERYONE WHO THINKS THIS TO TELL US THEIR CURRENT BUDGET FOR AD-HOC MARKET RESEARCH (IF THEY ARE A REASONABLE SIZE BLUE CHIP MULTINATIONAL) AND I WILL PROVIDE THEM WITH A 360 DEGREE SOLUTION THAT IS CHEAPER.

    Second, what sounds awesome as a conference sound bite (or written in a blog post, like this one) is technologically incredibly complex, massively big, and – unless you are the NSA, with virtually unlimited funds – extremely hard to do. To combine passive data for the same person reliably across multiple platforms at scale is something Richard Branson might consider beyond rocket science. NOT TRUE WITH HADOOP IN THE CLOUD ANYONE CAN AFFORD TO CRUNCH BIG DATA (THAT IS TERABYTES OF DATA FOR CENTS). THEY JUST NEED TO HAVE THE RIGHT PEOPLE WITH THE RIGHT SKILLSET. WE DIGITALMR CAN CRUNCH ANY DATA SIZE IN ANY PROCESSING TIME REQUIRED. DISTRIBUTED PROCESSING ALLOWS US TO ADD MORE MACHINES TO DEAL WITH THE SAME DATA SET (WITHOUT OWNING THE MACHINES)

    Just to provide a simple detail – managing a passive panel (not even census) requires constant oversight as people change devices, move to different states, and more. It also demands ongoing software development, as mobile companies change the way they do things; and it generates terabytes of data every day that need to be cleaned, processed, and made actionable by running them through ever-evolving taxonomies. And that is just for one platform; do this across digital and TV, and you are multiplying the complexities. ONGOING SOFTWARE DEVELOPMENT ABSOLUTELY. BEING CONTINUOUSLY IN BETA IS THE NEW WAY OF LIFE….WHY DO YOU THINK THAT THIS IS UNREASONABLE OR A PROBLEM. ALL THE PROCESSES THAT DESCRIBED HERE CAN BE AUTOMATED

    Third, we have not even talked about the definition of “census-level.” Does it mean all 200 million US adults? Is a sample of 20 million enough? In other words, do big numbers without actual sampling methodology truly “represent”? What if my 20 million represent a populous of mostly high-income people living in big metro areas? Is that “good enough”? THERE IS NO DOUBT THAT CENSUS LEVEL MEANS EVERYONE. I WOULD NOT CALL 20 MILLION CENSUS LEVEL FOR THE US. CENSUS LEVEL IN SOCIAL LISTENING IS WHAT WE DO CURRENTLY – NO BIG DEAL. NOBODY ADVOCATES THAT THERE SHOULD BE CENSUS LEVEL PANELS OR CENSUS LEVEL SURVEYS. INTERCEPTS WILL DO JUST FINE.

  2. Florian,

    Interesting article and interesting conclusions. There’s one other aspect that wasn’t mentioned is the government interference. At some point data privacy laws would get in the way of that level of collection.

    It’s an interesting thought experiment.

    Guy
    ProRelevant.com

  3. Seems as though you’re talking about gathering data for a macro view, but most businesses just want their own data. Luckily it’s becoming easier and easier to accomplish that with great accuracy with the rise of API connections.

Join the conversation

Florian Kahlert

Florian Kahlert

Chief Operation Officer, Cubeyou

Sign up to our newsletter

Don't miss out...
Get great insights content delivered straight to your inbox.
I agree to receive emails with insights-related content from GreenBook.


You can manage your email preferences or unsubscribe at any time. GreenBook protects your privacy under the General Data Protection Regulation.