For everyone who has ever discovered an unexpected consumer behavior insight, Robin Opie has a big one for you: “Clicks are not an effective predictor of purchase behavior.”
Robin is SVP of Analytics at audience measurement and media optimization company DataLogix. They have transformed their own business over the past few years to take advantage of the fact that the secrets to future purchase are not always where we think, and that a technological approach to identifying the right behavior drivers can boost productivity, scale and effectiveness.
“Consumers spend 35 percent of their time online, but 93% of buying is done offline,” he said in his keynote presentation today at NCDM13: Where Marketing Meets Big Data.
The problem is counter-intuitive to the great promise of digital measurement, he said. While 100% of click data can be captured, marketers are still having trouble finding consumers online. That means that 99% of those sales came from people who did not click, he said. “Much of the analytics work we do at DataLogix is to understand the relationship between the online exposure and the buying behavior,” he said. “For us, the Holy Grail is to identify the drivers across channels so that consumers can be recognized in any channel.”
We need to use better data to make better decisions about incremental revenue, not just the measured sales, he declared, so we must bridge offline data with the right online data. “Our goal is always to find the people who are likely to respond to an offer,” he said. And then, of course, we help marketers facilitate offers and measure sales results so they can repeat the process.
In the work of identifying the variables with the most significant impact on purchase behavior, Robin and his team rediscovered a principle that has been a hallmark of direct marketing for nearly a century: RFM (Recency, Frequency, Monetary) behavior. “We found that purchase data is really, really powerful,” he said. “About 90 percent of the variables that are meaningful in any model are purchase variables, and past purchase behavior is far better at predicting future purchases than clicks or proxy variables.”
DataLogix found that clicks were wrong half the time and immaterial the other half of the time. “Best case, you are targeting about two percent of your audience, and you may even be systematically pulling away from the other more valuable audiences,” he advised. He also cited studies from ComScore Facebook, and Nielsen also demonstrating that clicks are terrible at predicting purchase intent.
“There is an intersection between clickers and buyers,” he said. “It’s just not as valuable predictor of future purchase behavior as past buying behavior.”
On the other hand, households who will purchase in future are usually (but not always) similar to those who purchased in the past, he said. We identify that past purchase profile and model it to find others.
Time and again we see the incremental lift is dramatic, he said. Plus, these models can help you predict things like market share gain from future campaigns.
## ## ##
Join the DMA marketing analtytics community at NCDM15 at the fabulous Palmer House Hilton in Chicago in February 2015. More info and to register on our website.