The perils of buying data
How marketers can avoid the pitfalls of paying for 3rd party data
Data data everywhere
Every company knows they could be doing more with data. They’ve built complicated ERP systems to record every transaction in their business. They’ve spent decades capturing digital records of all their customer interactions. On top of that, they’re inundated with pitches from vendors with every flavor of 3rd party data. Vendors promising perfect demographic, contextual and intent data. If all the vendors are to be believed, marketers can find their exact customers, in any channel, right as they are deciding what to buy. There are two reasons why purchased data is less effective than promised — targeting cost and signal dilution.
The high cost of precision
The first reason for the lack of effectiveness of purchased data is cost. Companies selling data aren’t charities. It’s expensive to collect and store information on millions of people. Compounding the costs, most vendors aren’t the primary source for their data troves — they have to pay a network of publishers for the right to pixel their sites and collect user information. These vendors need to earn a return on their time and capital and thus have to charge marketers a premium for their data services.
For direct response marketers, performance and return on advertising spend are paramount. Most will pay for anything, including data, if they can achieve a commensurate increase in performance. The problem with paying for 3rd party data is picking. There are probably segments, among the millions of options, that will return more value than they cost. However, it’s difficult to predict in advance which segments will perform. There are hundreds of “Auto Intender” segments for a BMW dealership to choose from. One of these might perform well relative to cost, but maybe not. They might be better off targeting “High Income” or “Sports Enthusiast” segments, or something else entirely. It’s impossible to know. Most direct response marketers will have to suffer through months of testing poor performing segments before they find a few that justify the costs. Most performance-oriented budgets won’t allow for long periods of non-performant testing.
For brand marketers, there can be unintended consequences of paying up to increase targeting precision. Brand preference is the summation of a lifetime of messaging that slowly, over time, influences customer behavior. People aren’t loyal to Tide because they saw an ad on their mobile phone in their local Walmart. By the time a college student enters the market for detergent, P&G’s wants to have plied them with a tapestry of messaging to make sure the first bottle they reach for is red. Since licensing data is expensive and media budgets are fixed, increasing targeting precision results in less reach and fewer exposures. That’s why mass brands like P&G are choosing reach over precision, even on Facebook, which has the best targeting data anywhere.
The data lies hiding the truth
Facebook is the best case for data accuracy. It’s immense database is maintained by its users, it’s updated frequently and tied to a single identity (the user’s email) across channels and browsers. Most 3rd party data providers are bedeviled by the contradictory pressure to offer data accuracy and targeting scale. These pressures invariable lead companies that sell data to dilute their signal in search of scale.
In order to understand the contradiction that poisons most 3rd party datasets, it’s helpful to understand how these segments are created. Let’s construct a hypothetical “Auto Intenders” segment as an example. We might start by pixeling anyone who searches for new car price quotes on Kelly Blue Book. These people demonstrate strong intent to purchase a car and marketing to this group would likely be effective. The problem is that there are only a few thousand people a day that search for car prices. The supply is capped. Car-manufacturing advertisers have big budgets — they want to reach millions of people. It’s also a problem for our data-selling vendor. They get paid every time an advertiser uses their segment for targeting — small segments are bad business. So data vendors go searching for ever more sources of intent, with ever weaker signals. They add users that visit Car & Driver or that read blogs about cars. These segments invariably get bigger, bowing to pressure from both the marketer and the vendor’s business model. As the segments get bigger, the strong signals are polluted with less performant sources and performance suffers.
The alternative to purchased data
Most companies already have all the data they need. They just aren’t collecting, analyzing and activating it effectively. The best companies build comprehensive profiles of their customers. They are able to tailor messaging and experiences across channels down to the individual level. They send emails that users are excited to open and can predict how likely users are to open and engage with their emails before they’re sent. First-party data is powerful when used properly.
It will take a long time for most marketers to achieve that level of mastery over their own data. Until they do, they should not be wasting money buying data from others.
