5 min read

What are Ad Auctions? The Definitive Guide

Holly Shuffett
Holly Shuffett
Updated on
May 16, 2024
Intro to Ad Serving

Ad auctions and choosing the appropriate auction method are crucial components for publishers looking to build a successful ad business. But with so many options available, from various providers to methodologies, identifying the best ad auction strategy can seem overwhelming. 

To help you navigate these decisions and determine which auction model best serves your ad business, this article will define different auction methods, compare and contrast first-price and second-price auctions, and discuss when it makes sense to use ad auctions altogether. 

What are ad auctions? 

The Digital Marketing Institute defines ad auctions as, “The method used to decide the cost and placement of digital advertising on certain ad networks. A new ad auction occurs each time an ad is placed on an ad network.” 

Publishers run ad auctions as a way to maximize revenue. Auction logic is leveraged by an ad decision engine to select the ad that is most likely to deliver the highest payout. For example, if ten advertisers bid for one ad slot, the ad server would select the advertiser who is bidding the highest, similar to how an auction is run in the real world. 

Ad auctions step-by-step

A typical ad auction might look like this: 

  1. Choosing a platform/placement. First, you’ll have to choose the platform/ad placement advertisers will be bidding on. 
  2. Advertisers set their budgets. Advertisers determine how much they’re willing to spend daily or monthly. 
  3. Advertisers select keywords or audience segments. This step involves advertisers setting criteria that will help them target a specific audience. 
  4. Advertisers place bids. At this point, advertisers will set their maximum bids for the selected criteria. 
  5. Advertisers can monitor & adjust. After bids are placed, most advertisers will continue to review performance metrics to hone and refine their strategy. 

What is a waterfall auction? 

Waterfall auctions, also called daisy-chaining or waterfall tags, are defined by ClearCode as a process in which “...a publisher passes its inventory from ad network to ad network in descending order of importance until all impressions are (hopefully) sold.” Waterfall auctions are a byproduct of publishers’ efforts to eliminate unsold impressions, and due to a relatively simple technical setup, “waterfalling” is historically one of the most popular auction methods of choice.  

What is lottery selection? 

The most common ad selection method is “lottery” -- also called “even rotation” -- in which ads are randomly selected from eligible candidates, regardless of what the advertiser is willing to bid. In this process, each ad may also be assigned a dynamic weight to ensure it hits predefined goals, like clicks, impressions, or more.

Should I run auctions or lottery? 

Every publisher has different needs and commitments to their advertisers, and the strategy that works best for you depends on your needs and advertiser relationships. 

You might consider running auctions if: 

  • You have more advertisers than inventory. If you have, say, three advertisers all vying for the same ad slot, having them compete for it can drive more revenue for you. Similarly, if you only have three ad slots per page but nine advertisers, let them compete for those slots.
  • Your advertisers are asking you how they can drive more impressions. Offering advertisers a way to compete for more impressions through the form of higher bidding allows them to take control of their impressions.
  • You offer, or want to offer, a way for advertisers to set their own bids. With lottery selection, it’s odd to charge advertisers paying different amounts for the same inventory. Auctions allow advertisers to set their own bids.
  • You are not guaranteeing impression volumes.

You might stick with lottery selection if:

  • You guarantee volume. If you promise an advertiser 50M impressions for a fixed amount, you need to fill those. Lottery selection helps ensure that advertisers hit their targets rather than relying on what others are bidding.
  • You prefer fixed bids. Having advertisers set their own bids is complicated, especially for a fledgling ad platform. It’s simpler to launch with fixed bids (like $10 CPMs) that rotate evenly.
  • You have more inventory than advertisers. Like how more demand than supply can lead to a bidding war, the flipside could mean a race-to-the-bottom, with advertisers scrambling to pick up inventory at low bids. Using lottery with fixed bids helps by setting a “floor” amount you make.

Essentially, auctions can help maximize your revenue when demand outpaces supply, whereas lotteries are better when your pricing model is simpler, with fixed bids or guaranteed volume amounts.

What are the different ad auction types? 

There are two main kinds of ad auctions: first-price and second-price. Let’s explore each and compare. 

First-price auctions 

A first-price auction model is relatively straightforward: what you bid is what you pay. For example, imagine three advertisers bid $1.00, $2.00, and $3.00 respectively. $3.00 wins the bid, meaning the advertiser will then pay $3.00. 

Historically, first-price auctions have not been widely used across ad tech, but they have become increasingly popular in recent years. 

Publishers use first-price auctions because they: 

  • Offer greater revenue potential. You make what the advertiser bids, even if the next highest bid was substantially lower. 
  • Have become more popular. Google, for instance, migrated to a first-price auction model in 2021, upending the long held industry norm of using second-price auctions in programmatic advertising. 
  • Work better with header bidding. As ClearCode puts it, first-price auctions “...give buyers a better chance at winning the impression when header bidding is conducted, because instead of the winning bid from a second-price auction being pushed to the ad server, their true bid is the one that competes in the final auction.” 

Second-price auctions

A second-price auction model is when the winning bid pays $0.01 more than the second-highest bid. Using the same scenario in the previous section, if three advertisers bid $1.00, $2.00, and $3.00 respectively, although the advertiser bidding $3.00 would win, they would only have to pay the “clearing price” of $2.01 ($2.00 + $0.01) -- not their own bid of $3.00. 

At first glance, this model might seem inferior to first-price, but there are several reasons second-price models have long been the industry’s preferred choice. 

Publishers use second-price auctions because they: 

  • Are advertiser-friendly. Advertisers prefer second-price auctions because they won’t overpay for impressions. Lower prices will naturally lead to better downstream metrics too (cost per click, cost per action, etc.), which could keep them spending with you.
  • Ensure truthful bidding. If your advertisers know you do first-price, they may try to game the system via “bid shading,” where they estimate the market price of the impression, versus bidding what it’s actually worth to them. There’s a chance aggressive bid shading across the board could lead to less revenue than a second-price model (in response, publishers often implement bid floors). With second-price, however, there’s no reason to do bid shading; the auction will determine market value automatically.
  • Historically, are more common. Advertisers are accustomed to second-price platforms, but as noted above, this trend has been changing in recent years. 

What are different ad auction strategies? 

Whether you want to do first- or second-price auctions, utilizing auction logic for selecting a winning advertiser can get tricky based on your pricing model. The simplest method is to have all your advertisers bid via CPM (cost per thousand impressions). You’d then select the advertiser with the highest bid, maximizing your revenue.

But what happens if you want to offer CPC (cost per click) bidding, where advertisers set their desired CPCs? This adds a level of complexity, as the highest bidder doesn’t necessarily maximize your revenue. If someone is willing to pay $10 per click but nobody clicks on their ads, you make $0 from them. Meanwhile, if someone is willing to pay only $1 but gets clicks, you are better off selecting the $1 bidder.

With CPC bidding, therefore, the bid is important, but so is the expected click-through rate (CTR). And if you want to do cost-per-action bidding (CPA), you also have conversion rates to consider. These challenges have lent themselves to the emergence of auction strategies, like eCPM optimization and autobid. 

eCPM optimization 

As you build your ad business, you can find that writing a non-CPM auction algorithm yourself is difficult. The process involves complicated calculations of an expected-cost-per-mille, or eCPM, based on historical performance for every ad. 

Working with a third-party ad server can eliminate that work for you, or, if you want to build an in-house ad platform, integrating with Kevel can grant you access to that logic instantly. 


Autobidding, also called automated bidding, is another way advertisers can have more control, allowing them to specify a select few of their objectives and constraints, like which products they want to promote. Internal bidding bots will then do the rest through ML/AI algorithms to determine an optimal bid and dynamically adjust in response to changes in the bidding environment. 

Advertisers often have a specific goal in mind for their campaign’s success but may struggle to determine the optimal bid. Autobidding helps by taking the guesswork out of things, and according to Google Internal Data, more than 80% of Google advertisers utilized autobidding in 2021.

With Kevel’s Autobid feature, for example, advertisers only need to input their desired return on ad spend (ROAS) goal and maximum CPC bid. Kevel’s advanced ad decision engine handles the rest by automatically adjusting the CPC bid to optimize toward achieving the target ROAS goal. 

Key benefits of autobidding include: 

  • Better efficiency. Autobidding saves time and effort by automating bid adjustments, meaning less risk for advertisers of spending too much or too little. 
  • Better performance. Using historical data and predictive analytics, autobidding optimizes ad spend by calculating the most precise bid for each scenario. 
  • More flexibility. Autobidding lets you automatically adjust to market dynamics, while still managing multiple objectives and campaigns. 

What pricing model should I use for my ad auctions? 

Publishers generally offer three main pricing models for their direct-sold inventory: CPM, CPC, and CPA. 

Pricing Type Definition
Cost-per-mille (CPM) Total cost (ad spend divided by thousand impressions (mille is Latin for "thousand"). A $1 CPM across 1 million impressions would be $1,000 in spend.
Cost-per-click (CPC) Total cost (ad spend) divided by clicks. You would charge an advertiser, for example, $3.00 for every click you get for them.
Cost-per-action (CPA/CPL) Total cost (ad spend) divided by actions, conversions, or leads. For instance, you may charge an advertiser $50 for every lead you drive.
  • When to use cost-per-mille (CPM): For publishers, CPM pricing is the safest and easiest way to sell. As long as you have a pulse on your monthly impression numbers, you can easily make commitments based on that data. 
  • When to use cost-per-click (CPC): CPC is best for large ad platforms, due to its appeal to performance-focused advertisers. For publishers, however, CPC is a riskier model, since it introduces the unknown factor of click-through-rates (CTRs). CPC runs the risk of “wasting” impressions, should nobody click on an ad. 
  • When to use cost-per-action/lead (CPA or CPL): CPA or CPL is best for direct response advertisers. Here, advertisers pay only for a conversion event, such as a purchase or app download. This is even riskier for publishers, since you have to worry about both CTRs and conversion rates (CRs). If 100% of people click, but 0% of them convert, you make nothing, making this an uncommonly used pricing model.

Other questions to ask yourself when deciding your ad auction pricing model should include: Who are my advertisers? What are my competitors doing? How much risk and complexity do I want to take on? For a more in-depth exploration of pricing models, and how to determine which pricing model works for, check out our blog on it here

How do different publishers run auctions?


Kevel’s ad infrastructure APIs offer auction logic out-of-the-box. Depending on the tool, you can toggle first- or second-price, as well as set rules for auctions vs. lottery/even rotation. Kevel also offers eCPM optimization and Autobid features. 


Criteo runs Open Auction campaigns through first-price auctions on a CPC model. Advertisers can choose the bid they’re willing to pay every time a shopper engages with their ad and take part in the inventory auction to win the placement. 


Topsort offers auction-based sponsored listings, banner ads, and video ads monetization, and specializes in AI for Auction tools and New Age Infrastructure APIs


Google Ad Manager runs Open Bidding via a “server-to-server” integration between publishers and their demand partners. Once an ad request is generated, Ad Manager runs a unified auction to determine an appropriate yield, enabling third-party demand partners to compete for inventory in real-time. 


Moloco runs auctions driven by various ML techniques, which leverage event data provided by their advertisers. This technique allows them to build lookalike targeting models that show advertisements to other mobile app users in similar existing user bases as the advertiser’s target audience.


CitrusAd itself is an auction-based ad platform which allows advertisers to create advertising campaigns that are displayed on each retailer's e-commerce website. Advertisers bid to promote their products at the top of a retailer's website, and can track the performance of their products in real time.

What ad units work best with auctions? 

The ad units that work best with auctions are typically those with a high number of sellers, allowing for competitive bidding that can drive up prices. Auctions are not well-suited for high-dollar inventory, like carousel ads. Let’s take a look at some ad units that work well with auctions. 

Sponsored listings 

Sponsored listings, also called promoted listings or sponsored ads, are organic listings that are promoted or highlighted in some way, usually located at the top of search results. They are an innovative way to drive revenue from advertisers seeking more visibility, without having a detrimental impact on sales or user experience. 

Native ads 

Native ads are seamlessly integrated into your feed, homepage, search results, and more. Often, these native ads look identical to your standard content, expect with an added “sponsored” or “promoted” tag. Native advertising was a nearly $100B market in 2023 and is an innovative, engaging alternative to standard banner ads.

Banner ads

Banner ads are ads displayed on a publisher’s site in rectangular formats known as “banners.” This kind of display advertising can capture users’ attention and drive traffic to a brand’s site or help generate overall awareness. 

How do I get started? 

If you want to build auction functionality yourself, we recommend working with a data scientist to develop the algorithm needed for more complex auction decisioning. Your tech team would then need to incorporate this logic into your ad selection engine.

Of course, that path could take months and countless engineering hours. It takes time to customize auction logic and ensure it’s up-to-date with current tech innovations.

That’s why companies like DeliveryHero, Flink, and Yelp are increasingly turning to Kevel to easily incorporate auction logic into their in-house ad platforms. To learn more about how to do it too, contact us today.

All ad tech in your inbox

Subscribe to our newsletter to stay up to date with the latest news.