HOW DOES BIG DATA CHANGE MEDIA PLANNING AND MARKETING RESEARCH?
Denyse Drummond-Dunn posted a provocative blog on LinkedIn’s CMO Network, “Is There Still a Need for Marketing Research Departments?” She raised a number of interesting questions. The one that I found most interesting was “Am I capable of accepting that true insight development doesn’t come from one study or database, but from information integration of multiple sources?”1 In my opinion, the biggest impact of large-scale data is directed to the marketing research organization. Not that it gets downsized, but that it has a more important role in sourcing and organizing big data.
In the past, marketers invested much more research into obtaining “reported observations” using marketing techniques. Unfortunately, a reported observation is a recollection of the facts, so it loses accuracy, and it is also dependent largely on how the question was asked. The good news is now we have a lot more “real observations,” representing what actually happened. So, how do organizations get access to these real observations, and how much investment is needed before the data can be translated into marketing actions?
Take the example of media research conducted by syndicated media companies like Nielsen. Much of the media research in the past was based on small samples. Nielsen ran a panel of television viewers and another panel of shoppers, and used their data to project the results to the population using census and other secondary sources for extrapolation to the population. Today, with the availability of cable viewership data from cable set-top-boxes (STBs), cable operators are aspiring to supply their television viewership data to marketers. In the United States, their first major attempt was Project Canoe, which was formed as a consortium of six cable operators. This was in response to a Google-Echostar partnership, which used satellite viewing on the Dish network. When it comes to advertising, Google is not shy about stating its ambitions. “We are confident we are going to revive the television advertising industry" says [Google TV’s Vincent] Dureau, “by bringing new advertising to it" Already, Google is trying to make TV ads more relevant, easier to target, and cheaper to deploy. As a result, Google thinks it can attract more ad dollars from smaller businesses that may not have been advertising on TV before.2
Unfortunately, the Canoe project failed in meeting its original goals. Six years later, the effort was scaled back to providing interactive advertising for video-on-demand. While Canoe was not able to get all its constituents on the same page, it did in some ways show proof of concept. It released the results of a study it jointly conducted with the American Association of Advertisers, in which a panel of 4,200 cable subscribers revealed increased product acceptance when shown interactive ads from brands like Honda, Fidelity, GlaxoSmithKline, and State Farm. According to the year-long study, 19 percent of adults 18-49 said “yes” to interactive offers, while 36 percent expressed a likelihood to purchase.3
The interactive video market has grown in the meantime, leading to yet another avenue for media and advertising research. Unlike linear television (which is shown on airwaves), interactive television, often termed “nonlinear” is well instrumented. Video content managers keep track of viewing details for each of their subscribers. As the nonlinear content is becoming increasingly mobile, the location of the subscriber may no longer be static. Last, but not the least, interactive viewing is more often done by an individual as opposed to a household. In my own house, interactive video has significantly reduced our traditional television viewership. The Canoe project finally found its sweet spot, and has survived and thrived as a supplier of interactive television research data.
The online viewership of data and video is closely scrutinized. Data management platforms (DMPs), such as Blue Kai, Adobe, Aggregate Knowledge, CoreAudience, Knotice, nPario, and X Plus One, track and analyze a fair amount of data from Internet viewers. Marketers are very interested in understanding a cross section of viewers, a holistic viewer- ship data, including social, mobile, display, and search. Potentially this data can then be combined with the delivery of messages via targeted advertising or in other ways. With the rise in the mobile platform for viewership, this data can also be correlated with location data to understand and mine geographic differences.
A number of research organizations have started to collect and analyze shopping behavior. For example, I mentioned Four Square and Slice as data sources in chapters 2 and 3, respectively. A marketing research organization can use their analytics to understand shopper behavior using Four Square to map shopping, and Slice to track buying. As these organizations gather critical mass, they are able to provide valuable insight into the customer’s buying process. In addition, most marketers have their own loyalty cards, which provide additional buying patterns, although focused on the brands supplied by the marketer.
Let us consider the difficult task faced by the marketing research department. By subscribing to a couple of market leaders in media research, a market researcher can easily access a vast pool of data about their customers. However, the data may have serious biases. Let me take a look at the situations I discussed earlier, in which wireless service providers and cable companies were able to offer and package mobility patterns and content usage to the retailers. If a retailer purchases a lot of smartphone, Wi-Fi, and cell tower location data from a wireless service provider to analyze traffic in a mall, the population is most likely smartphone users and not a random population. If a marketer uses audience data from a digital STB, it excludes a section of the population that is on analog STB devices. The first task for the market researcher is to identify the bias, if it is likely to skew the results. In the above examples, both samples may represent a bias toward affluent consumers. If the marketer is using these sources to model the price-sensitive buyer, the results are not going to be accurate. The second task is to use targeted research to fill in the gaps. Often, hypotheses regarding biases can easily be validated or rejected by doing directed research on the component that is underrepresented by big data. The marketer can sample portions of the observations to test the hypothesis. These examples carve out a new role for the marketing researcher—the one who integrates a variety of sources of data and adjusts the data before it is misused.
Integration of data is the next challenge. Chapter 3 showed a set of overlapping big data sources. As a researcher starts with the census data, each new data source provides additional attributes, which can be overlaid on top of census data to project a view of the population, assuming the data is properly aligned across big data sources.
Big data is also changing the measurements and KPIs in media planning and marketing research. In the good old days, advertising was focused on getting as many eyeballs (“reach”) on an advertisement for a given budget and as many times as possible (“opportunities to see” or OTS) to facilitate memorization of the messaging. Since most of the programming was linear, media research organizations kept track of reach for various programs by different targets, and a media planning algorithm could maximize reach and OTS for its target segments for a given advertising budget. Click-through rate (CTR)4 and cost-per-click (CPC) emerged as the first set of measures for online advertising. Google and others have used CPC as a measure for bidding on advertisement placement.5 While it is the first step in the evolution of advertising effectiveness, CTR often does not result in sales. Most Internet users employ those sites for a variety of information retrievals and do not have an explicit goal or desire to click on advertisements.6 Also, clicks are only related to browser activities, while media exposure that led to the click could be inspired elsewhere. A more convergent scenario is one in which someone may be watching television and browsing for advertisers’ products on a second screen.7 The action on one medium is dependent on exposure on the other media, and the advertising allocation must sense and correlate multiscreen viewing by customers. While there is a tendency among marketers to isolate the effectiveness of each channel in order to decide on budgets for each, the overall experience from a customer perspective is holistic.
Fortunately, the raw data supporting a gigantic set of events is gradually becoming available. The cable operators are working toward audience measurement data that can record audience viewing for specific programming and commercials. Internet viewing and related commercial exposure data is available from the data management platform suppliers. Mobility data can tell us the action in the form of physical shopping, while online shopping is available from online content analytics. In a study conducted for IAB France and SRI, the consulting firm PwC has recommended a five-level advertising effectiveness model using display, actual exposure, interaction, browsing, and engagement. PwC further specifies nine categories of indicators for measuring online performance using display, conversion, traffic, interaction, subscription, media, distribution, ROI, and posttests.8 These measures are only associated with online advertising. A comprehensive set of measures and KPIs for multiscreen media research is still evolving.