Commentary

How Advances In Machine Learning Are Making CPG Marketing Near-Real Time

The ad biz is moving closer to real-time marketing thanks to advances in machine learning.

Data that previously would have taken weeks or months to be processed, analyzed and, most important, activated are now being activated in near-real time.

The latest move forward comes from NCSolutions (formerly Nielsen Catalina Solutions), the joint venture of Nielsen and Catalina Marketing that enables marketers to correlate actual in-store purchasing data with marketing inputs -- media, targeting, creative, etc. -- to understand what’s driving the outcomes, for good or bad.

To date, the data have been somewhat latent -- enabling marketers, agencies and the media industry’s sell-side to explain what happened after the fact, but not in time to make “in-flight” changes that might actually make a difference during a campaign's run.

“In the past, a campaign would complete and you would get you results six weeks later; now you can get it weekly,” NCSolutions Chief Research Officer Leslie Wood says of the breakthrough, which is being rolled out as a new product aptly dubbed Sales Lift Metrics.

Wood, a renowned ad industry media researcher (and winner of last year’s ARF Erwin Ephron Demystification Award), says the breakthrough is only possible now because of improvements in data processing and data servers.

In fact, she says, NCSolutions tried applying machine learning to the same kind of data five years ago and failed. But it didn’t give up, and the new product reduces the latency from weeks to one-week intervals.

The data primarily represent the kind of consumer packaged goods products and brands that are bought in stores, and the pilot customers typically look at the data in batches of weeks to understand longer-term patterns, but “this is as close as you can get to real-time in the CPG” world, Wood says. 

“As an advertiser, our budgets are often evaluated based on our ability to drive sales of our products. Historically, as an industry, we have been limited to using top-of funnel-KPIs [viewability, reach, etc.] to optimize our media in-flight. That left us always asking ‘Which metric is most correlated with incremental sales?,'" points out Paul Gelb, head of programmatic and social at big pharma marketer Bayer. Gelb notes that the new data enable Bayer to evaluate the performance of its media, creative and other elements on incremental sales while the campaign is live.

Wood says the data is good enough to automatically inform and adjust programmatic media buys on the fly. The next phase is to integrate it with dynamic creative optimization systems so that copy and messaging also are tweaked in near-real time, to maximize incremental sales, she adds.

While a number of marketers have or will be piloting the new data, Wood says the customer base most likely to embrace it quickly is the supply side -- media companies that are eager to demonstrate how their inventory generates incremental sales lift, or how it can be tweaked to a brand’s or agency’s advantage.

Next story loading loading..