One of the most important online advertising trends of the past couple years is the rise of algorithmic custom bidding products: Google’s Performance Max (PMax) and Meta Advantage+ Shopping Campaigns (ASC).
These tools plug directly into an advertiser’s data warehouse, SDKs and servers with the promise of producing real business outcomes, not flimsy programmatic metrics like click-through rate.
The growing popularity of platform-based custom bidders has helped normalize the idea of AI-based bidding, giving a boost to a small market of programmatic-focused non-platform customer bidding startups that existed before the walled garden versions.
Scibids and Chalice Custom Algorithms are the two most notable examples. Both are pursuing the benefits of first-party data and business outcomes, but without the black-box decision-making you get with Google or Meta. PMax and ASC advertisers, for example, don’t know when and where their ads ran, or even what creative was used.
But can Scibids and Chalice build this new category before algorithmic bidders become yet another platform plaything?
AdExchanger took a look under their hood.
Scibids
Scibids was founded in 2016 before there was a formal “custom bidder” market.
Every advertiser wants to generate outcomes specific to their business and has unique data sets to determine those outcomes, said CEO and Co-Founder Rémi Lemmonier. “And yet, until recently they’ve been using one-size-fits-all bidding algorithms” that don’t take those elements into consideration.
Part of what Scibids does is to take an advertiser’s purchase and profile data and turn it into a proxy or custom model of an advertiser’s desired metrics.
Not all companies have sufficient purchase or profile data to create custom bidding algorithms – but there are other useful datapoints.
Site engagement data, in-store foot traffic or info pulled from an ad (like how many people completed a video and/or clicked to the site) can be used as proxies for valuable actions taken on the way to a conversion.
“You don’t need 1 million conversions to predict 1 million conversions,” Lemmonier said.
The third-party custom bidder market relies on modeling because a custom bidder requires a strong dataset to start with and a livestream of data to keep it sharp. PMax and ASC use an advertiser’s own data but also rely heavily on their own platform’s first-party data. Google and Meta have more than enough, so advertisers simply add their data to the pile.
But Scibids only has the advertiser’s data and data from third-party sources, such as a location data provider.
Still, there are multiple ways that custom bidders can differentiate from the big guys, including through pricing, control and transparency into how the AI works and the bidding decisions it makes, Lemmonier said.
PMax may be a “custom” bidding algorithm, but it’s not really customizable. There’s no way for advertisers to control campaigns as they can with non-walled garden tech. PMax advertisers can’t finetune targeting based on specific demographics, time of day or by media formats.
As long as they continue to see positive ROI, advertisers may not require more transparency from PMax or ASC. But those platform custom bidders are plagued by doubts and suspicion because their bidding decisions are inscrutable. Does PMax benefit the advertiser primarily, or does it benefit YouTube, Google Search and all the other Google products? It’s impossible to say from the outside.
But Scibids provides an AI that “is really independent,” Lemmonier said, and optimizes “100% to the advertiser.”
DoubleVerify, which acquired Scibids in August for $125 million, was attracted to its independence.
“We have a common philosophy around independence, which is being outside of the buying decision,” DoubleVerify CEO Mark Zagorski told AdExchanger.
Scibids does make bidding decisions and it charges a percent-of-media fee, which makes it more like a DSP than DoubleVerify has been in the past. But it isn’t the media-buying platform.
And since Scibids isn’t buying media, it also isn’t bound by log files. It can be used to model audiences on walled gardens or other channels that are closed to programmatic.
Zagorski said that DoubleVerify’s expansion plans for Scibids are to apply its bidding tech beyond the open web across social channels, such as Meta and TikTok, as well as streaming platforms like Netflix.
Although custom bidders are “an emerging application in the programmatic ecosystem world,” Zagorski said, “there’s a significant opportunity to move well beyond the open internet.”
Chalice Custom Algorithms
Chalice, which was founded in 2020, is the other big fish in the still small custom bidder market.
Although, it’s Chalice that often turns away potential customers “for being too small,” CEO and co-founder Adam Heimlich told AdExchanger. Heimlich added that Chalice charges a $2 CPM fee, “with the intent of turning away people who spend $1 on media.”
There is also a threshold in terms of what constitutes “enough data” to make it worth using a custom bidder, he said. Take the Democratic party, for example, which is one of Chalice’s clients. Whereas a state level campaign in Georgia with tens of millions of impressions is big enough, a campaign in Massachusetts would be below the threshold.
But it’s Chalice’s CPM fee structure that’s one of the main distinctions between it and Scibids, which is a pure SaaS business.
Venture capitalists typically prefer subscription software, which they consider more reliable and desirable than earning an ad tech CPM fee, Heimlich said. But Chalice’s only VC investors are from ad tech: The Trade Desk’s VC arm TD7 and the fund AperiamVentures.
“The Trade Desk flies in the face of that wisdom, as they don’t charge a SaaS fee – they charge a media fee, and they’re worth $50 billion,” Heimlich said.
Chalice isn’t pure ad tech, though. Like Scibids, there’s a great deal of manual work and programmatic expertise involved, which can fall under an ad hoc consulting fee.
“The vision is to move toward self-serve and for Chalice to do less service than today,” Heimlich said.
In the meantime, since Chalice launched in 2021, the time to onboard a client and create a custom algorithm for them has decreased from weeks or months to one day or even 10 minutes.
Advertisers start with an off-the-shelf bidding algorithm based on their goals, such as a political campaign or a CPG brand that run post-campaign surveys to measure recognition or brand lift. The most popular is an AI Allocator, Heimlich said.
Take Hershey’s, for instance, a Chalice client that is laser focused on selling all of its Halloween candy. By Nov. 1, Halloween packaging is outdated, which means the brand needs to allocate spend specifically to sell out where it has Halloween candy in stock.
If candy is flying off the shelf in New York, a typical programmatic optimization would involve pushing spend into that region to follow the demand. Retail networks do the same.
But it would make more logical sense for Hershey’s to encourage people to buy Halloween candy in areas where inventory is unsold.
Custom bidding enables smarter targeting, but also offers transparency into and control over post-campaign analytics. Advertisers can finetune their campaigns, see where ads were served and see both which creative units and which publishers drove value for the business.
Those details are blank for PMax or ASC buyers, Heimlich said. But perhaps the biggest value of custom bidding tech is the ability to understand performance by publisher or media channel so as to align targeting strategy with creative.
“The market is bifurcated,” Heimlich said. said. Chalice’s market, he said, is the growing minority of brands that want a greater level of control.
