Challenges of measurement in a digital age

Media research has been roiled the past few weeks by a number of interesting developments. First Rentrak and comScore announced a merger clearly aimed at challenging Nielsen, which has long held an effective monopoly on TV ratings. Just hours later, Nielsen said it would roll out its combined TV-online ratings by year’s end, something media people have been clamoring for for months. It comes at a time when media research has never been so complicated, yet also never been so important. New ways of collecting data offer the promise of better targeting and more effective media planning, if only they can be harnessed. And wearables and smartphones have incredible promise for achieving something buyers have always hoped for: Apples-to-apples ratings comparisons across media. But all this will take a lot of time, which is frustrating for buyers, who feel as though they’ve already been waiting too long for research to catch up to changing media habits. Pete Doe, chief research officer at programmatic sales platform clypd and former senior vice president of data science at Nielsen, talks to Media Life about the future of measurement, how research has changed in recent years, and how the Rentrak-comScore deal could push Nielsen.

 

Why did you decide to join clypd? What will you be doing there and how will it differ from your duties at Nielsen?

Clypd is a great company-we’re doing groundbreaking work, and I feel very at home with the culture and people. I see my main responsibility at clypd as ensuring the quality and client value of the data and algorithms in our programmatic platform.

I had the same focus on clients, data and algorithms at Nielsen, but my role here is further downstream-using data rather than supplying it-and that’s exciting.

At clypd we’re working with data sets that I helped create in my previous job.

 

What do you see as the biggest issue in TV ratings right now?

The fragmentation of viewing across platforms and content providers is the biggest challenge.

The industry wants integrated measurement that includes content consumption across all platforms with sufficient precision to look at small segments of the population and detailed viewing events-and all overnight please!

To achieve this requires integration of data set such as panels, big data sources and digital census data-and this creates many challenges for research suppliers trying to put all the pieces together, as well as challenges for data users trying to keep up with the increased complexity of the data and research architecture.

I should also add that these are global issues, not just U.S.

 

What are media people most interested in learning about TV measurement that they’re not getting right now? Why?

Aside from the cross-platform discussion, another dimension is that of impact-based planning and execution, the idea that you would buy on outcomes delivered by a campaign such as purchase, request for information, visits to a car dealer, rather than just audiences.

Whether that is achievable as an integrated industry-wide currency play is debatable. It is more likely to be a custom capability in my view.

At clypd we are focusing on helping media sellers and buyers reach more precise brand targets, going beyond demographics and using secondary data sets with more detailed characteristics to make ads more relevant and effective in delivering the right audience.

 

Nielsen has gotten a lot of flak over the years for moving on its own timetable for changes, rather than that of its clients. Do you think this reputation is fair? Why or why not?

I think the assumption in the question is that Nielsen clients all have the same timetable and that is not the case.

When I was at Nielsen, I recall some clients telling us to go quicker while others were telling us to go slower.

Any currency supplier faces the same issue-there is so much riding on the numbers that any change needs to be examined in great detail by all interested parties.

I always remember what a senior industry figure in the UK said in a heated meeting many years ago discussing a new measurement system: “With currency measurement, clients want progress, but no change.”

 

Many media people would like to see ratings that measure across a variety of media and could make for apples-to-apples comparisons for, say, TV vs. out of home. Do you think that could ever be achieved? What would have to happen to get those in place?

There are data sets that measure across a variety of media, of course-MRI, Scarborough and Simmons for example, and integrated data-sets and analysis systems that enable cross-media planning.

The apples-to-apples comparison is open to debate as the effectiveness of the different media varies by campaign, category, brand, creative etc.

Looking to the future, could there ever be a measurement that is sufficiently comprehensive and precise to capture all media exposure? If it happens then I would expect smartphones and/or wearable technology would be at least part of the measurement solution.

 

What’s the most misunderstood area of media research right now? Why?

Relevant to what we’re doing at clypd, the word “programmatic” seems to be one that causes a lot of misunderstanding.

I think many people assume that it means “addressable,” whereas it has a broader and different meaning. We describe programmatic as workflow automation and data-driven decision making.

Improving the placement of ads through algorithms and advanced targets benefits buyer, seller and the viewer.

 

What’s the most important area of media research right now? Why?

I think the opportunity and need to integrate traditional research sample data with big data sources is the most important area.

I can’t think of a medium that isn’t touched by the digital revolution, and in all cases some digital exhaust data could potentially be harnessed to better understand the medium, whether it’s TV, digital media-of course-radio, print or outdoor.

That is leading to new research methods with statistical models that are daunting to many researchers.

I think we have to remember that all research data and methods past, present and future are statistical models of the population, and if we bear that in mind we should be open to considering these new methods and data, while maintaining rigorous evaluation criteria based on sound statistical principles as we have in the past.

Related, the digital revolution is transforming buying and selling everywhere, and that’s gaining pace in our space with programmatic TV.

 

This is obviously a time of great change for viewing patterns. How has this impacted media researchers?

It’s made our jobs harder!

But also more interesting-the world has changed more than most people would have expected, even in the last five years. The main change I have seen is the emphasis on data science as a must-have capability.

Research has always been about data, of course, but the digital dividend of big data sources has created a whole new set of computing and statistical techniques and skills that are now central to research. When I started in research, statisticians and IT were seen as a quirky but necessary element in a research organization and most “real” researchers were happy to be seen as distinct and separate.

Now researchers are all learning [data science tools] R and Python.

 

What do you think of the Rentrak-comScore merger? Can they become a challenger to Nielsen?

I’m sure they would consider themselves as already challenging Nielsen, particularly in local TV measurement and digital, and I think both Rentrak and comScore probably realized that they needed to merge to remain competitive with Nielsen.

In the end, of course, any currency is determined by agreement between buyer and seller, with the MRC as independent arbiter of quality. No measurement system is perfect, but as long as people understand the pros and cons, they can make an informed decision as to which solution best works for their business.

So, while most people would see Nielsen’s people meter measurement as the state of the art, it isn’t affordable in smaller markets, and clients want ever bigger samples in the larger markets and nationally. Rentrak offers large samples but is constrained by return path data availability and provides measurement of devices rather than people.

 

 

 

 

 

Tags: clypd, comscore, nielsen, nielsen ratings, pete doe, Q&As, rentrak, research