By Alexandra Bruell. Source: WSJ.
For companies that sell consumer products like soap or beer or cereal in a variety of retail stores, it’s always been tough to quickly discern whether their advertising directly leads to a sale.
Despite the promise of sophisticated digital ad measurement, many packaged-goods companies are still reliant on an antiquated and laborious process to measure whether their ads drove people to buy their products in stores.
A combination of inadequate technology and comfort in old marketing metrics are responsible for the status quo, but that may be changing.
Typically, Constellation waits up to 12 weeks for a data vendor to send reports—which can cost between $45,000 and $65,000 apiece—to get information on how its digital ads for certain brands were driving purchases of alcoholic beverages across stores. The lag time made it difficult to discern the effectiveness of a digital ad campaign until after the fact.
So, Constellation decided to set up a trial with Black Box Wines, which was diving into digital advertising for the first time. The marketers for the wine brand found that with a tweak to existing technology from an ad tech firm and a data vendor willing to provide a weekly influx of purchase data, the company could optimize its digital video and display ads in the middle of a campaign, versus long after its end.
“What we really wanted to do was better understand how our marketing was impacting the bottom line,” said Jaymie Schoenberg, vice president of marketing for Black Box Wines. “We wanted to push the envelope on more real-time optimization. As we thought about diversifying our marketing mix and bringing in a mix of TV and digital during key selling periods, it felt like the right time.”
Packaged-goods companies typically have relied on marketing metrics like awareness and brand lift, versus sales, because of the lag time in getting sales data from outside vendors like retailers.
But Constellation’s latest experiment serves as one example of progress in a still maturing digital ad market.
The company, which shifted about a quarter of its TV ad spend to digital channels, tested the campaign during the holidays, from November to January.
One of Constellation’s data providers—the group previously known as Datalogix, which was absorbed by the Oracle Data Cloud—agreed to offer up collected purchase data every week or so while the campaign was running. With the help of programmatic consultant Labmatik, Constellation also enlisted the help of Turn, an advertising technology company with a data management platform. A DMP is a kind of software that can collect and analyze data used to send ads to specific groups of people online.
Turn then “hacked its system” so it could reformat and ingest the data from the Oracle Data Cloud. The technology update enabled Turn to match data showing the anonymized identities and behavior of the people who saw the ads with the Oracle data showing the anonymized identities associated with households and people who bought Black Box in stores.
About every two weeks, the Black Box team got a report from Turn showing more anonymized detailed information about the households and individuals actually seeing the ads and buying the product. Based on those findings, Black Box got a sense of which kinds of people and households were responsive to the ads and which ad formats and frequencies were working. The brand could then adjust its ad plan accordingly and target people with similar attributes.
For example, in the midst of a campaign, Black Box could send more video ads to women of a certain age and income level—attributes similar to the people and households that bought the wine.
The experiment “has probably taken us more toward the Holy Grail everyone is looking for,” said Brenda Tuohig, vice president and general manager for consumer products at Oracle Data Cloud.
Ad clicks have been a “suboptimal mechanism to optimize media tactics” since so many consumer-packaged-goods products are bought offline in actual retail stores, she said. “We were able, in this case for the first time, instead of optimizing toward a click, to think about purchase and transaction.”
In mid-November, Constellation established how many people and households were seeing the digital ads and also buying the wine offline. In December, when it began optimizing ads based on learnings from the purchase data, the number of people and households who saw the ads and bought the wine increased 49%, said Karena Breslin, vice president of digital marketing for Constellation. From December to January, average monthly offline conversions increased 55%.
The company also learned that females aged 25-54 were converting at a faster rate, meaning they saw the ad and then purchased the product. “Previously, we would have said we thought [the customer makeup] was more equal in terms of male versus female,” said Ms. Schoenberg.
It also learned which ad formats and sites were driving sales, as well as the best timing and frequency of the ads, she said.
“The ability to optimize against offline sales data while in-flight during the holiday campaign was what made [the effort] unique,” said Ms. Breslin. “In the past, we were unable to optimize while in-flight.”
If Oracle Data Cloud starts offering this kind of service to more brands, the company could charge clients in a variety of ways—on an impression basis, a percent of media spend or a flat fee if it’s part of a larger “enterprise” relationship, said Ms. Tuohig.
“We’re looking forward to further rolling it out across other brands and platforms,” she said. Much of Oracle’s offline purchase data comes from customers who are enrolled in loyalty programs.
“CPGs sell their products to grocery stores, and they lose the visibility into the data and sales,” said Scott Hagedorn, CEO of Omnicom media agency Hearts & Sciences. This “relatively new” kind of data capability is “especially interesting to them.”
For Black Box Wines, this trial offered insight into its target consumers and gave it a jump-start on planning for its next ad campaign.
“We’re definitely looking to find more ways to get real-time data,” said Ms. Schoenberg. “One thing we see as an additional opportunity is how can we potentially shorten the data cycle even further.”