Every single time a person opens a website or application that is ad-monetized, a new auction is initiated to determine which advertiser will be able to serve the user an ad. In each auction, all advertisers who desire to show their ad to the user are are stack-ranked according to various factors, and the winner of the auction then proceeds to show the user their ad.
While the inter workings of digital advertising auctions are highly dynamic and full of complexities, this article will hopefully shed some light on the process for those less familiar with it, in order to help advertisers learn how to better optimize their campaigns.
Before reading further, please note that the exact methods by which auctions operate are not fully known to public parties (or even most folks who work on the auctions) and this article should be taken as a guideline for understanding applicable concepts, rather than stone, granular truth.
This article breaks down the ad auction dynamics into four distinct sections listed just below, and serves up some links for continuing the learnings. While this article is useful for understanding both web and mobile auctions, it focuses on the mobile advertising perspective.
- How the auction happens: step-by-step
- How ad rank is calculated
- How the price paid is determined
- The event optimization layer
- Topics for further reading
How an ad auction occurs: step-by-step
As stated in the introduction, an auction is initiated when a user (hopefully not a bot user, which is another story) opens a website or application, and an ad inventory spot (an impression) appears on the user’s screen. This auction kicks off the following sequence:
- All advertisers on the ad platform are assessed for eligibility to show the user an ad.
- Eligibility is determined based on whether the advertiser’s settings enable them to serve the user an ad. Settings can include operating system, geographic targeting, negative keywords, user install status, budget availability, and others depending on the network.
- Each eligible advertiser is ranked from the best ad rank to the worst.
- The best-ranked advertiser wins the best ad placement (usually the placement highest on the user’s screen), the second best-ranked advertiser wins the second-best placement, and so on until there are no more placements available for that user.
- The user then chooses to engage or not with the advertiser’s ad.
- Engagement can mean many things depending on the ad platform and the campaign type, but generally engagement means the user clicks, installs, or performs in-app event. In some platforms such as Facebook, it can also mean clicking “I don’t want to see this ad anymore,” or (reportedly) slowing down while scrolling through an ad in a news feed.
- For CPM (cost per thousand impressions) bidding, the advertiser is charged just to show the ad. For CPC (cost per click)/CPI (cost per install)/CPE (cost per event) bidding campaigns, the advertiser is only charged if the user completes the targeted event.
- The results of which advertiser wins the auction, and more importantly the user’s engagement inform the stack rank of the next auction, and thus which advertiser wins the next user’s ad impression. In this manner, auctions are a self-virtuous cycle favoring the winners, all things equal. This fact can make it challenging and expensive to climb out of a low ad-rank hole, once an advertiser has fallen into one.
How ad rank is calculated
In order to determine which advertiser wins the impression, advertising auctions use ad rank to stack-rank advertisers and their ad campaigns. In most advertising auctions there are three main components to ad rank:
Ad rank = (historic/predicted CTR/CVR * bid) / (ad quality/relevance score / maximum ad quality/relevance score)
- The bid is how much the advertiser is willing to pay to show a user an ad, or else how much the advertiser is willing to pay for the event they are bidding for. The bid is the easiest thing for the advertiser to control, and while unbounded can often loses effectiveness once it is raised to a certain level, without improvements in historic/predicted CTR/CVR.
- That is, raising bid from $10,00 to $100,000 is unlikely to yield a different result in auctions won for an advertiser, as $10,000 is already likely to be higher than virtually every other bid, meaning that the advertiser gains no incremental benefit from raising the bid at this level.
- The historic/predicted CTR/CVR:
- Predicted click-through-rate, AKA CTR (in some cases predicted event conversion rate AKA CVR is also used) is the most well-kept and influential components of ad rank. While less complex auctions use historic CTR, advanced auctions bring to bear all of their capabilities in determining which user is highly likely to engage with each advertiser’s ad.
- Ad platforms are keen to predict whether a user will engage with an ad or not, because while an ad network may be able to make money per impression, the network can charge the advertiser far more money on a click than an impression, more money on an install than a click, and more money on a purchase than an install. Additionally, the network will be better able to retain its end-users and advertisers (and their budgets) if it is able to find the users most likely to engage with ads.
- Quality/relevance score is a defense system put in place to weed out bad actor advertisers whose products are harmful to the user experience. While this component is bounded (typically 1-10) and can thus be overcome if either numerator component (bid or historic/predicted CTR/CVR) is high enough, it nonetheless has a weighty impact on how many impressions an advertiser will win and also the price the advertiser pays.
Not only does the advertiser’s rank score determine whether an advertiser will win an impression, but it also influences (or is correlated with factors that influence) the price paid by the winning advertiser.
In many cases there is also a minimum ad rank or floor bid necessary for an advertiser to win an auction. This prevents an advertiser from winning an impression at a rate lower than the network's cost (in either operational costs or user experience costs) of monetizing the user’s impression.
How the price paid is determined
The amount that the advertiser pays is calculated based on the advertiser’s rank, as well as the competition.
Advertising auctions operate based on a second price auction, where the winner does not pay the price they bid, but rather an amount just higher than what the next best-ranked advertiser would pay, often referenced as “one penny more.”
This means that advertisers at the upper levels of the ad rank stack (especially in auctions where there are fewer total advertisers) are often able to influence what prices the winner pays. In the PPC world this would often lead to bidding wars, where the advertiser in position #2 could push up prices on the advertiser in position #1 to drain their budget without paying any more themselves, except in cases where advertiser #2’s ad rank was higher than advertiser #1’s.
By similar means advertisers are also able to form a moat and defend against lower-ranked advertisers by raising bids to levels that retain the winning bid and still only paying enough to beat the second-best ranked advertiser. This was initially highly prevalent in the PPC world of Apple Search Ads, due to the fact that only one ad at a time is shown. Yet this strategy can easily backfire if the second-best ranked advertiser also decides to raise their bid (or improves their historic/predicted CTR), resulting in an upward price war damaging for both parties.
While ad auctions began by focusing on a simple outcome (i.e. showing a user an ad impression or getting the user to click an ad), modern advertising auctions now enable advertisers to optimize for more and more complex outcomes (e.g. getting a user to view at least 10 seconds of a video or make a purchase).
In order to allow advertisers running different event optimizations to bid for the same user’s impression, each advertiser’s bid must be converted on the back end to a standard amount, which allows advertisers bidding different amounts for different events to participate in the same auctions. Bids are most commonly converted into a CPM amount to make this possible.
For example, an advertiser bidding $.5 for a click may have their bid normalized into a $15 CPM and an advertiser bidding $100 for a purchase may have their bid normalized into a $120 CPM. In this case, the advertiser bidding for a purchase outcome is the most likely to win the auction, due to having the higher CPM bid. Yet, this advertiser would not want to pay this amount to show their ads to every user as only a small percentage of total users will end up making a purchase; deeper funnel bidding optimizations therefore focus on showing ads to a smaller group of users, targeting only those who are likely to drive that deeper funnel outcome.
This is also where the predicted CTR/CVR also comes into play, as each user may have a different propensity to complete different events. For instance, for a user who is unlikely to click an ad but may watch 10 seconds of an ad, if a single advertiser bidding for 10-second video views enters an auction filled with only advertisers bidding for clicks, if all else is equal, then the 10-second video view-bidding advertiser would win this auction due to having a higher predicted likelihood of viewing 10-seconds.
Topics for further reading
Here are a list of other topics that can help expand your understanding of how digital advertising works:
- Other Incipia posts
- 5 diagrams explaining how UAC campaigns work (including auction/ad targeting mechanics)
- Why mobile marketing benefits from algorithmic marketing (i.e. event bidding)
- Facebook lookalike saturation part 1 (why does it occur)
- Facebook lookalike saturation part 2 (how to address it)
- What’s the difference between web and mobile marketing?
- Viewable impressions (not every impression served is created equal or has a chance of actually earning a click or conversion - more common in the web world)
- Official documentation on how auctions work
- Google’s chief economist explains the AdWords search auction (Hal Varian)
- Facebook how auctions work (help article)
- Header-based bidding (Eric Seufert)
- Remnant advertising (Andrew Chen)
- Real-Time Bidding/RTB
That’s all for today! Thanks for reading and stay tuned for more posts breaking down mobile marketing concepts.
Be sure to bookmark our blog, sign up to our email newsletter for new post updates and reach out if you're interested in working with us to optimize your app's ASO or mobile marketing strategy.
Incipia is a mobile marketing consultancy that markets apps for companies, with a specialty in mobile advertising, business intelligence, and ASO. For post topics, feedback or business inquiries please contact us, or send an inquiry to firstname.lastname@example.org
- A/B testing
- android development
- app analytics
- app annie
- app development
- app marketing
- app promotion
- app review
- app store
- app store algorithm update
- app store optimization
- app store search ads
- apple search ads
- aso tools
- client management
- coming soon
- facebook ads
- google play
- google play algorithm update
- google play aso
- google play console
- google play optimization
- google play store
- google play store aso
- google play store optimization
- google uac
- google universal campaigns
- ios 11
- ios 11 aso
- ios 14
- ios development
- itunes connect
- limit ad tracking
- mobiel marketing
- mobile action
- mobile analytics
- mobile marketing
- play store
- promoted iap
- promoted in app purchases
- push notifications
- search ads
- universal app campaigns
- universal campaigns
- user retention
- ux design