When Goodwill wanted to drive foot traffic to their stores as well as create awareness about their larger mission of providing job training and support to local communities, they turned to AdTheorent, a leading provider of predictive digital ad technology.
Formed in 2011, AdTheorent is a machine learning-powered, targeted advertising provider. Their technology uses countless data signals that exist in the digital ecosystem to effectively and efficiently target digital ads to consumers based on their clients’ goals.
ClickZ spoke with Jim Lawson, the CEO of AdTheorent, to discuss Goodwill’s recent mobile campaign, including gaining a better understanding of the Cost Per Incremental Visit (CPIV) model which requires that advertisers only pay for incremental store visits (e.g., “visitations”).
Goodwill’s goal in launching a campaign with AdTheorent was to drive visits to their brick-and-mortar stores as efficiently as possible. They learned about AdTheorent through a partnership with Ad Council.
Ryan Kuhm, Goodwill’s Interim CMO, writes, “Goodwill Industries International is very focused on finding the most efficient and effective means to reach targeted consumers with our message, and have been focused on the advances in geotargeted and tracked advertising technologies. When Ad Council approached us with this opportunity, and we saw AdTheorent’s proposed capabilities, we knew it was something we wanted to explore.”
AdTheorent’s Jim Lawson explained that they are plugged into a vast amount of ad inventory via ad exchanges which facilitate the buying and selling of inventory across multiple ad networks.
“What the ad exchanges allow us to do is hyper target users on an impression-level basis using non-sensitive personal information from the bid stream and other sources,” explains Lawson. “We leverage data such as operating system, phone type, time of day, publication type, weather patterns, geographic location, and other signals. Our system then correlates variables that existed when conversions occurred and determines what variables yield visits to Goodwill stores so it can optimize delivery of ads to those people more likely to convert.”
Typically, ad targeting is based on retargeting or audience-based targeting and not based on data-driven, real-time signals. But AdTheorent’s technology makes this possible by tying ad variables to users’ mobile data. For example, if there are people standing within a mile radius of Goodwill and someone is on the NYTimes app, AdTheorent can purchase the available ad space in real-time via the exchange marketplace and serve up the ad to the person who is within a certain radius of the store.
For Goodwill, the ad unit itself was a display ad that linked to the Goodwill website. The ads appeared on mobile apps with available inventory available throughout the length of the campaign. Once an AdTheorent campaign is launched, it takes a few weeks for the technology to learn what works best based on the campaign goals (in this case, store visits or “visitations.”)
Says Lawson, “In Goodwill’s case, the system optimized ad delivery towards the best performing creative sizes with the 320 x 50 having the highest percentage of visitations. The machine learning model optimized toward that creative type.”
The system also recognized that users connected via wifi were more likely to convert, Verizon users were more likely to convert, and users at home versus those at work were much more likely to convert to a visitation. They also found that Monday and Tuesday were the best converting days.
“These are things you can’t anticipate in advance and plan for,” says Lawson. “Our system is about using big data and the vastness of the data and letting the system tell you what drives conversions and then optimizing towards those variables.”
AdTheorent utilized its cost per incremental visit (CPIV) ad pricing model for the Goodwill campaign. The CPIV pricing model guarantees that brands only pay for incremental foot traffic resulting from ad exposure. AdTheorent works with a third-party measurement partner that uses technology which enables them to determine whether an ad drove someone to visit a store.
“There are digital IDs that were exposed to the ad and we share that with our partner,” explains Lawson. “Our partner has a very large panel and when there’s overlap they can determine the visitation of their panel and extrapolate the conversions of the exposed panelists to the broader audience to determine the incremental lift caused by seeing the ad.”
AdTheorent is paid per incremental visit based on the third-party report. “If we don’t drive visits, we don’t get paid,” says Lawson. “The key to being able to do this is in making sure you have enough scale and enough data, so you can be confident that the models will determine or detect correlations that will drive performance.”
AdTheorent has rules around which types of campaigns are eligible for CPIV pricing. They get reporting throughout the campaign from their visitation partner and provide a wrap-up report to the client at the end of the campaign which breaks down things like day of week, time of day, top performing segments and demographics.
Sample AdTheorent Report
Lawson explained that they negotiate the CPIV in advance with each client and recommend a minimum six-week duration for every campaign, so the software has time to learn what works.
“Models, by their very nature, need to learn,” Says Lawson. “Oftentimes in our campaigns, the first couple of weeks are the learning period though this could be a shorter or longer period of time.”
In Goodwill’s case, the combination of machine learning and a hyper-targeted geofencing approach drove good results in terms of getting people into their local Goodwill store.
As mentioned above, one of the key goals of the campaign was to generate awareness about Goodwill’s mission. Goodwill did this via the campaign creative units using the message “Shop Goodwill, Bring Good Home.”
The ads linked to the Bring Good Home page on Goodwill.org where users could read about Goodwill’s mission and view a variety of content including a video and a graphic.
Goodwill was happy with the results of this campaign which far exceeded their expectations. AdTheorent delivered a 470% lift in incremental visitation at nearly 80% less than the contracted CPIV. While these results were more than satisfactory, they also learned a lot from AdTheorent’s data-driven, machine learning approach to serving ads.
“The learnings will inform how we continue to advance our outreach beyond this single campaign,” explains Kuhn. “This campaign is refining the expectations we set for all of our geotargeted efforts. We saw intel that was counter-intuitive from what we know. That is that the data showed an increase in visitors on Monday and Tuesdays versus weekends when more opportunistic shoppers visit Goodwill stores.
We learned the true cost and impact differences between a standard geotargeted “store visit” and those that represent a standard lift as well as a behavioral lift. These insights are extremely helpful in attribution of our marketing efforts, which we have not had previously, and we plan to explore in much more depth.”
The post Case study: How Goodwill uses AdTheorent to drive visits to stores appeared first on ClickZ.Reblogged 11 months ago from www.clickz.com