Amazon, eBay, Etsy, and many other marketplaces have proven that sponsored listings inserted into search results can drive revenue and separate them from competitors.
But building such an ad solution is complicated. How do you connect your ad server with your search/recommendation engine? How do you account for complex ad targeting around search term permutations and range filters? Does it even make sense to hack an ad server into a search engine?
Kevel has recently launched a new use case of our technology called “Auction-as-a-service”, which helps brands address these questions and more easily insert personalized sponsored listings into their search results, thereby helping them drive revenue and improve ad relevance.
Brands that want to insert sponsored listings into search results face a conundrum: either spend a year plus building it from scratch, or use a third-party tool that’s inflexible and not equipped to handle a myriad of search term combinations at scale.
In the latter scenario, they have to effectively hack their ad server into a search engine...which isn’t ideal.
This is because standard ad servers, whether built in-house or outsourced, generally employ key-value pair targeting, whereby you create an ad targeting certain keywords, and if the ad request contains that keyword, it’s eligible to be shown.
For instance, you could create an ad with the logic “Is eligible to be shown if the user searches for "2021 Honda Accord"”.
Then, when a person makes a search, you would pass their search term (or “keyword”) in the ad request. The ad server would cross-reference this phrase with every ad’s keyword logic. In the above example, that listing would be eligible if the user has searched for "2021 Honda Accord" (though not variations like "Honda Accord 2021").
There are a few issues with this set-up, though, for search-based ad targeting.
One, trying to account for every possible search permutation means adding potentially hundreds or thousands of keywords to the ad. It’s likely a simple misspelling means an ad is glossed over. And the more keywords the decision engine has to work through, the slower the ad response times.
Two, what happens when your ad server picks a sponsored listing that the search engine doesn’t think should be displayed? Do you insert it or not? By having both services recreate the search process, you run the risk of irrelevant listings being displayed to the user, creating poor user experiences.
Using a feature called AdQuery, we’ve enabled a way for brands to bypass Kevel’s targeting logic and go straight to the auction and tracking mechanics.
In this scenario, rather than asking both your organic and paid engines to sort through all available listings, you first ask your search tool what listings are eligible to be shown. Generally this is tracked as a SKU (stock-keeping unit) or an internal product ID.
You would then send these eligible IDs to Kevel in the Decision API request.
From there Kevel would look just at the ads associated with those IDs and run the auction against that set, incorporating logic around ad prioritization rules, revenue optimizations, pacing, budgeting, frequency capping, and so on.
Depending on how many placements you need, we’ll send back one or multiple IDs that you should promote (and where), along with the tracking URLs needed for impression, click, event, and conversion reporting.
There are few ways that this "auction-as-a-service" tool helps brands who are looking to build or scale their sponsored listings ad platforms:
Even better, AdQuery enables a whole bunch of additional functionality too, all of which help improve the user’s ad experience and help publishers drive the most revenue from their ad platforms. We’ll be outlining these use cases in future posts.
Learn more about employing 'auction-as-a-service' using Kevel’s Decision API. We’re happy to set up a quick consultation to walk through your specific needs and discuss the best path forward.
