Mindshare Chief Data Officer Rolf Olsen and Oleg Korenfeld, EVP of ad tech and platforms at Spark Foundry, will speak at AdExchanger’s upcoming PROGRAMMATIC I/O New York conference on Oct. 25-26 in a presentation titled “Data Accuracy and the DMP.”
Partnering with a data management platform doesn’t count as a data strategy, said Rolf Olsen, chief data officer at GroupM agency Mindshare.
“An internal data strategy is often the missing piece,” Olsen said. “You may have a bunch of CRM data, but you also have to think about how accurate and reliable it is, and that’s often about a lot more than just signing a contract with a DMP.”
Inaccurate or stale data can lurk in any file – and that becomes an even greater risk when partnering with data providers.
Advertisers and their agencies need to interrogate the data and the third-party data providers that serve it up, even before testing it.
But there’s no simple way to validate either first- or third-party data sets without actually doing the test, said Oleg Korenfeld, EVP of ad tech platforms at Publicis media shop Spark Foundry, which recently rebranded from Mediavest.
“We can’t know what makes one audience pool perform better than another until we try it out,” Korenfeld said. And testing can be expensive because it requires advertisers to spend money on media anyway.
So before the trial-and-error part, Korenfeld approaches data partnerships with a lot of skepticism. He and his team ask a long list of questions. Even a marketer’s own CRM file should be questioned and probed for weaknesses.
“If it’s our client’s data, we want to know how it ended up in their CRM before onboarding it,” Korenfeld said.
And third-party data providers get the brunt of the interrogation: Where did the data come from? How fresh is it? How often does the provider refresh it? Why did certain people end up in a particular segment? Was the segment created probabilistically or deterministically?
At Mindshare, for example, Olsen has come across “fishiness” around geolocation, which has so many providers “with varying degrees of accuracy.”
“Say we’re talking about a mall,” Olsen said. “That means you need to account for both the physical location and the fact that there could be several floors in the building. That’s why it comes down to understanding where a signal is coming from.”
Data strategy is a combination of asking the right questions to vet partners and conducting an ongoing process of test-and-learn.
“As you move from exploration to testing, you often come across challenges with the data you may not have expected, which is why you have to experiment to find the nuggets of data that are most appropriate to the client,” he said. “There’s a lot of work that goes into making data as efficient as possible.”
Getting dirty with the data is expensive and it’s not easy, but that’s the only way to home in on quality before moving onto the next step of joining data together from different databases to create a fuller view of the customer, Olsen said.
“Even if your data strategy is sound, you increasingly need to think about how to aggregate and connect your data,” he said. “It’s not cheap, but the juice is well worth the squeeze.”