Programmatic bidding algorithms aren’t always aligned with campaign goals, so agencies are relying on AI tech to create algorithms tailored to their campaigns instead.
Butler/Till ran a display ad campaign on behalf of a financial services client between August and September last year that utilized tech from SWYM.ai, a company that provides programmatic decisioning tools to aid in deal curation.
The agency used SWYM.ai’s SCaLE optimization tool, which stands for Smart Curation and Learning Engine, to overcome some limitations baked into the campaign strategy and legacy ad tech.
The tool allowed Butler/Till to evaluate more potential impressions within the campaign’s narrow geotargeting than the agency would have been able to otherwise, said Ryan Lammela, group director, channel activation at Butler/Till. And it curated the impressions that were most likely to convert into a private marketplace sold through Index Exchange while optimizing that PMP on a daily basis.
As a result, the brand saw a 56% increase in conversion rate and a 26% decrease in cost per conversion, compared to a control portion of the campaign that did not use SWYM.ai’s tool.
The tool also reduced the number of domains on which the campaign ran by 52%, which helped Butler/Till cut out nonperforming inventory and avoid made for advertising (MFA) sites, Lammela said.
Avoiding MFA was particularly important because these impressions don’t perform as well as non-MFA impressions, he said. Plus, agencies can’t always trust the conversion data they get from MFA sites, where users are more likely to accidentally click on the profusion of ads on the page.
Curating performance
In other words, Butler/Till’s main motivation for using the SWYM.ai tool was to cut programmatic waste by focusing only on highly performant inventory and high-value leads.
Agencies need guidance when targeting high-performing inventory, because DSPs are trained to send buyers bid requests for the impressions they’re most likely to bid on – not necessarily the impressions that will perform best for the campaign, Lammela said.
So Butler/Till used SWYM.ai’s tool to curate a PMP that only contained ad impressions for users that were highly likely to convert. To build these PMPs, the tool examined sell-side optimization signals provided by Index Exchange, as well as buy-side signals derived from the DSP, in this case Google DV360.
Attribution data from users interacting with the ads was pulled from Google Marketing Platform’s Floodlight conversion and event tracking. And data about which publishers served high-performing ads – and which ad formats performed best – came from Index Exchange.
The goal of the campaign was to raise awareness for the financial services brand. So the campaign KPIs included site visits to the brand’s owned and operated page, as well as leads generated when users provided their contact info, Lammela said.
The SWYM.ai tool was able to identify which bid requests most frequently led to landing page visits and lead generation, said Andrew Altersohn, co-founder and president of SWYM.ai. It then packaged lookalike impressions into a PMP that could be updated on a daily basis, based on these optimization signals.
“We’re identifying aspects of the bid requests that have performant characteristics – anything from domains to geos to ad sizes to device types to channels,” Altersohn said. Once it’s identified these characteristics, the AI tool can create a PMP of similar impressions via an API integration with the SSP.
Curating at scale
Of course, working through one SSP and limiting the number of publishers on which the ads ran isn’t exactly a recipe for scale.
Plus, the campaign was targeted to users within tight geographic areas around specific physical locations, Lammela said. Restricting a programmatic campaign to a handful of ZIP codes limits the campaign’s potential reach and the amount of targeting parameters that can be reasonably applied, he added.
But the SWYM.ai tool helped Butler/Till achieve greater reach by finding bid requests within those tight geos that the DSP may not have surfaced otherwise, Lammela said. Plus, because the tool is passing optimization signals back and forth with the SSP, the SSP is better able to send bid requests that align with what the buyer is looking for.
“Any tool that allows us to have more bid requests that are being evaluated [within those geos] becomes extremely helpful,” Lammela said. Butler/Till’s goal was to level up the amount of impressions it was getting in these tight areas in order to have a greater pool of performant inventory to pull from.
In that sense, deal curation and AI are more than just hype for Butler/Till, Lammela added. They’re helping the agency take back control from legacy algorithms when it comes to how to spend clients’ budgets.
“We can take a little bit of control back from DSPs that are usually pretty heavily focused on QPS, and not necessarily giving us opportunities that are going to convert, but opportunities that we’re going to buy,” he said.
This approach is also a way to avoid the race to the bottom that has biased buying platforms toward surfacing cheap reach sourced from MFA and other shady publishers.
“We want to scale, too,” Lammela said, “but when we can have scale with performance, obviously it’s better.”
