How the Ad Industry Encourages Poor Data Quality
Next time you’re selecting targeting data for a digital ad campaign, pick segments that have won awards for quality, received five-star user ratings, or have rave reviews by data aficionados. Makes sense: Data decisions guided by true quality indicators tend to yield better results.
Unfortunately, you can’t do that—there’s no such indicator.
Audience segments are sold through a data store akin to a modern retailer. A single DSP may offer over 130,000 data segments through its data store without providing a single clue whether the merchandise is good or bad. Imagine if data sets were physically packaged and sold in a brick-and-mortar store.You’d find yourself standing amidst aisles and aisles of brown paper grab bags, trying to decide which generically packaged product you should buy to drive the growth of your business.
While a high percentage of third-party targeting data comes from the centralized repositories of highly respected brands like Oracle, Liveramp, and The Trade Desk, there’s no clear evidence that the data has been vetted for accuracy by these companies or their data partners. Claiming that they “must remain independent,” data store proprietors intentionally provide neither guidance nor differentiating details and leave the choice up to the user. Consequently, marketers end up having to decide between long lists of options that look alike.
Data store operators actually benefit from putting every possible product on their virtual shelves without any description, discretion, or liability because the more data they sell, the more money they make. The less information they provide, the fewer reasons buyers have to rule out segments.
This is a problem for data buyers who want to make informed purchases. Data is compromised by factors like vendor sources, processes, and packaging techniques, but in today’s data store, there’s no way for a quality-conscious buyer to properly gauge or compare data quality. Worse, given little opportunity to distinguish their products on the shelves, data vendors have little incentive to focus on accuracy or any other differentiating quality benefit.
So, the industry is reaping what it has sown: a high-demand, low-quality marketplace in which “buyer beware” can’t apply because buying decisions can’t be informed, and quality has no comparative value. In the data store, products are only distinguishable by price, scale, and product name, creating additional problems.
Price: When you’re looking at a list of seemingly identical segments that match your search, the cheapest segment likely holds the greatest initial appeal. As a result, data providers, motivated to be the low-price leader, are less likely to invest in quality assurance. In the end, it turns out, cheap data can be a very costly deal—low-quality data is now a leading cause of wasted ad impressions.
Scale: Data vendors make money each time an impression is served to a device targeted by their data sets. At the same time, marketers have scale requirements for their campaigns. These two motivations often work symbiotically to make segment size an overriding factor in purchase decisions. The biggest segment often wins. This could encourage segment stuffing (adding as many devices or cookies as possible even at the expense of confidence they belong). But the sheer scope of some super-sized segments should prompt probing questions about how the vendors gather and shape the data. In one data store, a segment includes 96 million devices of consumers actively in the market for a Mercedes Benz—but a total of only 17 million cars were sold in the U.S. of any type in 2017. Yet another targets 6.9 million people in Oral and Maxillofacial Dental Radiology. Do these numbers sound right to anybody?
Product Name: Ad platforms typically present segment options alphabetically. Given the lack of opportunity to meaningfully distinguish themselves, vendors use the old Yellow Pages tactic of naming their products so they top the list. In one data store, there are over 200 data providers listed and nearly 60 percent start with a number, a symbol, or begin with one of the first nine letters of the alphabet. For example, data vendor Optimus replaces the “O” with a zero at the beginning of its name. Vendigi starts its segment names with a “#” sign.
The point is, digital advertising is an incredibly sophisticated industry, yet when it comes to selling in the data store, vendors lack incentive and opportunity to focus on quality. Instead, they’re restricted to sales tactics—none of which are good for buyers, brands, or the industry.
It’s time for the industry to get serious about data quality and expect more from each player in the data value chain. Here are a few steps that would significantly improve data quality and buying experiences:
Data repository operators should be required to vet the accuracy and size of the segment they’re representing. That’s not to say that options should be limited. Buyers should be free to balance the trade-offs between scale and accuracy however they choose—but they should be presented with enough details to make informed purchase decisions. Currently, they’re only getting information on half of that trade-off (scale).
Third-party data-buying interfaces (DSPs, data stores, etc.) should be expected to provide comparable evidence of the quality of the segments they offer. Moreover, they should naturally assume a sense of responsibility to do so regardless of market expectations.
Data stores should facilitate and expose reviews from past users to help guide future data buyers. Community feedback would allow the data store to maintain neutrality while improving discoverability, enhancing ease of use and encouraging quality.
If data stores, vendors, and platform providers don’t take these steps on their own, marketers should demand data accountability. Ad data is now a $20 billion industry. It’s not unreasonable to expect some degree of segment discoverability and comparison within data stores. At the least, it’s essential that segment attributes like quality scores and proof of accuracy are revealed wherever data is sold.
Jake Moskowitz is head of the Emodo Institute, a dedicated organization within Ericsson Emodo wholly focused on the research, education, and resolution of data concerns that mobile advertisers face.