Marketing Data: Number Crunchers Need Communication Skills, Too
Sometimes data crunching is as much about knowing how to communicate it as it is the actual data work. For Keith Gooberman, VP of trading and platform operations at MDC Partners-owned Varick Media, the ability to translate his work to business execs was key to professional advancement.
"My career really exploded when I started to be able to talk to people about it," said Mr. Gooberman, who oversees the media-trading department and platform managment team for the agency trading desk.
A lifelong New Yorker who grew up on the upper East side of Manhattan, he studied mechanical engineering at Union College in Schenectady. Then, realizing he'd have to move away from the city to pursue a career in engines or rocketry, Mr. Gooberman ended up in the online-ad industry instead, getting his start as a temp at Conde Nast Digital working for the publishing giant's director of finance. His job was to determine how to sell and price digital ads, identify revenue streams and conduct rate card analysis and financial reporting.
He moved on to niche publisher Glam Media, where the ad impressions were whizzing by. At Glam, he said, he was "dealing with 10 million and 20 million impressions that are passing through the exchange everyday."
At Varick, Mr. Gooberman has assembled a team of around 8 people -- economists, math majors, computer scientists and engineers -- to handle data management and media buying, and assist in building the firm's proprietary platform technology. "We're building a UI that we can actually use."
Much of what he does involves acquiring and vetting data sets to help inform ad targeting. For example, he uses weather data for some clients. "It's crucial to know if it's snowing if you're working with a food seller," he said. "My job is to buy the inventory against that."
Ad Age: What do you wish marketers would understand about what data scientists do?
Mr. Gooberman: This is a great question, but I actually would prefer this was addressed in the reverse: I wish data scientists would understand more about marketing. Data scientists know how to pull statistically significant samples of aggregated information from large data sets. True data scientists are not well-versed in advertising terms or what each piece of data actually represents. They often need to be assisted in extracting valuable insights from a data set. This is not because they don't know how to build a regression algorithm, or how to parse the data, but simply because they are not aware of what is important for a marketer.
Data scientists should understand how data impacts their business. Data scientists enjoy parsing enormous amounts of data very quickly. They value the speed at which they can ingest and compute massive amounts of data. Furthermore, data scientists like to make grand statements based on sample indicators they pull from large data sets. In advertising, the best insights are often minor alterations in trends which occur over long periods of time (and take time to see due to their nuanced nature). Advertising It is more about the art of storytelling than it is about having the fastest processes.
Ad Age: You often assess new data sets from companies that are just starting to monetize their data. Who's out there selling data to Varick you didn't expect to be doing so a year or two ago?
Mr. Gooberman: While all of the [personally identifiable information] is stripped and audience segments are aggregated through online behavior, I am surprised to see financial companies with spending data (credit cards) offering their consumer's data to buyers on the exchanges and networks. Companies who handle customer reviews as well as website and blog 'commenting' platforms are offering their data, but since it is broad and cookie based, this is not shocking. They are just online tech companies trying to monetize their products.
Ad Age: What's unique about the information that credit card companies like Mastercard offer and how do you vet the quality of data offered by vendors?
Mr. Gooberman: Mastercard first removes all PII. Then they aggregate the users with basic purchase styles which are built from their purchase history, but are not brand specific. This allows them to remain privacy compliant while giving us some useful data.
The process of vetting data from all vendors is similar. It usually involves placebo testing as well as optimizing it as a separate strategy, as opposed to pure prospecting. With remarketing, as with any third party data, the more recent the user-indicated interest or behavior, the more compelling the data. Recency is the strongest indication of performance (ie: if I'm looking at sneakers now, I'm going to buy them in the next three days).
Ad Age: What's the coolest or strangest type of data set you've ever worked with and why?
Mr. Gooberman: I love seeing loyalty card data actually impacting CPG-style campaigns. While there exist several degrees of separation when online CPG campaigns correlate directly to in-store sales, these can be interesting and fun to analyze (assuming they cover enough scale). Also, a personal favorite has always been how weather data impacts buying habits (especially food). I saw enough data (especially around college markets) to craft an argument claiming the combination of age, demo (M18-54), geo, and hometown athletic team jersey colors had an impact on domestic used car color decisions and price.
Ad Age: Digital targeting capabilities have become so diverse and complex. Is there anything you wish you could do now that you cannot?
Mr. Gooberman: I believe it is time schools got involved. Major educational institutions can begin to segment their student bodies by study major without infringing on privacy. For example, FIT students could see more retail ads, and I don't think their students would mind.
Also, location-based services like Foursquare or Yelp should begin trying to serve their data in target-able segments. Imagine you are Target, and you are opening a Target in a new location – offering mobile coupons via geotargeted ads around the new locations only to users who constantly visit Walmart and Kmart would be extremely effective in generating in-store traffic. And that is just one idea. Also, nobody has done a great job tying in-store sales to cookie data. This would need a loyalty card provider, a matching engine, and a DMP.
Ad Age: What educational fields of study and professional backgrounds help develop the best data people?
Mr. Gooberman: I am biased, but I prefer individuals who studied engineering but spent their professional careers in computer programming. This means they have an engineering degree (not computer science) but only work in computer science environments professionally. This is naturally somewhat counter intuitive, but this means they were taught to build physical objects in school, but now use computer languages to handle the construction of said objects.