AI-Based Bidding is the Future of Advertising

Have you ever visited a marketplace and engaged in haggling only to realize in hindsight that you paid too much? Buyer’s remorse doesn’t feel good, but don’t be too hard on yourself if you are not trained or skilled in such negotiations.

If you have experienced a similar regret after paying too much for digital media, you should explore the benefits of AI-Based Bidding (ABB) for your next advertising campaign.

AI-Based Bidding is an advertising mechanism for bidding at a variety of different prices for the same campaign line item, prices that differ depending on time of day or week, device type, media type, and a variety of other factors to achieve better outcomes for the client.

Digital Service Platforms (DSPs) bid in real-time on behalf of advertisers for the right to show ads. Within milliseconds, an auction takes place among contending advertisers for a particular position on a webpage or within an app. The winning advertiser is the one who bid highest for the ad slot. These days, the winning advertiser will likely be required to pay an amount equal to that winning bid (a so-called first price auction), though some smaller percentage of the time the winning advertiser might only be required to pay the amount of the highest losing bid (a second-price auction).

Either way, a relevant question is how much an advertiser should bid. Bid too low and the probability of winning the auction will be too low. Bid too high and you might win the auction but pay more than necessary. It’s like balancing on a seesaw, but not as fun!

How much to properly bid to achieve that balancing act is something that requires intelligence and automation. The Intelligence aspect requires an understanding of which bidding situations are expected to result in better or worse outcomes. For example, it might be more expensive to bid during evening prime time than it is to bid at 1 a.m., but there might be far fewer digital IDs to target at 1 a.m. Win probability is somewhat a proxy for inventory availability, and inventory might be more available for the desktop ads in the middle of the day, but ads delivered via mobile in-app might be more available on weekends and holidays. Restrictive geo-targeting conditions might make inventory scarce for a specific campaign, whilst corresponding nationwide inventory remains plentiful.

Trying to do all that manually is too much for most AdOps teams, so instead they typically use a “flat CPM” strategy to bid at a constant value, no matter what. But that leaves lots of money on the table. Automation is required to consider all the myriad different context feature combinations. It’s simply too much work for a human. At Catalina, we found tens of thousands of $100 bills just lying on the theoretical sidewalk, ready to be plucked! But to collect those $100 bills, we needed an AI-Based Bidding (ABB) solution to automatically decide whether to bid high or bid low, depending on the values of the different bid context features such as time of day, day of week, media type, whether there’s been a win or not yet for this digital ID, and many others. These collective context features define opportune moments at which to bid higher and focus on winning or bid lower and focus on cost savings.

We use machine learning algorithms to bid higher when the combination of context features is most likely to lead to outcome KPIs of interest to our clients (e.g., ROAS), and bid lower when the combination of context features is less likely to do so. Doing this requires digital measurement and machine learning algorithms that embody the following concept: if I’m going to bid highly, then I’d like to do so in situations for which the viewing of that ad is likely to convince the viewer to purchase the promoted product.

Earlier, I mentioned bid context features, which sounds a lot like contextual targeting. That’s not an accident! Contextual targeting is how marketers intelligently deliver digital media in the absence of persistent identities. As privacy concerns erode the digital media industry’s ability to target based on persistent shopper identities such as MAIDs, cookies, and IP addresses, the need to be able to bid appropriately when the identity of the viewer is not known will only increase.

A key aspect of digital contextual targeting is to take into account the nature of the website or app where the ad is viewed: just as it is important to know whether the visitor to the store is tarrying in front of a collection of clothing, lawnmowers or spices, so too is it important to know whether the website in question is focused on clothing, equipment or cooking.

But beyond just the website and app characterization, other important context features are the time of day, the geography, the media and device types, and other characteristics of the bid. The collection of all of these context features inform the who, what, where, why and when of the bidding opportunity. ABB can take all these factors into account, using them as substitutes for identity and establishing the solid foundation upon which true contextual targeting – minus identity - can be built.

Nothing is so constant as change. Yesterday, a line item might have been pacing just fine, but today it’s really struggling. Maybe the marketplace suddenly got more competitive due to normal Q4 seasonality. Or maybe a slew of news sites were blocked following political turmoil. A good ABB solution must roll with the punches and dynamically adjust prices in response to changes in win probabilities or other measures of inventory availability or scarcity. And the ABB solution needs to adjust as the impressions and measurement signals become clearer as the campaign progresses.

ABB is an in-flight optimization capability - constantly adjusting to changing conditions. Note that ABB and contextual targeting rely on detailed understanding of micro-markets, not on broad-brush statistics. ABB and contextual targeting look at each line item separately and understand what is going on today rather than relying on aggregate statistics constructed from many unrelated line items.

Looking at all the micro-markets and dynamically changing bids every day requires software. It is too much for a human to do manually. But the profitability gains are there. AI-Based Bidding is the wave of the future.

To learn more about this groundbreaking advancement and how it can transform your CPG marketing efforts, reach out to Catalina today.