How Meta Ads Auction System Works
Meta ads resources, why the system cheats and what to do about it
I was recently asked how Meta ads auction system works.
I (embarrassingly) stumbled over an answer that basically equated to “uh, we just need to make good ads.”
My initial response felt like it lacked a clear explanation of the underlying mechanics of Meta’s algorithm even though I do believe that creating great ads is the practical goal.
This post aims to address that more clearly. It also includes some commentary on why the Meta system will cheat if you let it and what to do about it.
Lastly, I’ve also included a list of helpful resources for further deep dives on Meta’s ranking system at the bottom of this post. Enjoy!
Ad Engine incentives
The engines want to prioritize ads based on the advertisers that they think have the best chance to hit their campaign goals (IE: drive a user to purchase).
The circular logic here is that if the advertiser hits its goals, then it is more likely to keep spending with the engines. And, if we keep spending with the engines, then the engines make more money.
The engines use expected value calculations to determine which ads to serve over others (more on that below).
How the engine calculates expected value
According to Meta’s documentation, there are three key factors that go into calculating the expected value of an advertiser, which determines the ad auction ranking:
Bid
Ad quality
Estimated action rate (click, email submission, purchase, etc).
I like the simplified calculation that the MobileDevMemo blog presented for purchase event optimizations as a double click:
Expected value of ad impression (used for prioritizing ad ranking) = bid x probability of completing goal event
In MobileDev’s calculation, they wrap ad quality and estimated action rate up in the probability of completing goal event metric.
I usually translate the probability of completing goal event into an expected conversion rate (eCVR) because it’s easier to wrap my head around.
In essence, if I am able to create high quality ads (ad quality) and improve the conversion signal back to Meta either through landing page optimizations, promo offers, better signal tooling, etc., then I can improve my expected conversion rate (probability of goal event) that Meta is using to determine ad rankings.So then my focus of work becomes a) creating better ads and b) improving conversion signal back to the engine.
The combination of these two will improve rankings.
A quick note on better ads…
Creating “better ads” is a bit subjective but I like Barry Hott’s explanation of what great ads are. He mentions that they should be:
Relevant (target the right audience)
Potent (hit a key pain point / say something worth saying)
Engaging (drives an emotional response)
More on that to come in a later post.
Okay, wrapping up on expected values…
You may notice Meta spend ramp up during times of high conversion rates (IE: Black Friday Sales) and ramp down when we see lower conversion rates (IE: website bug / conversion pixel issues) independently of ad changes.
This is because the Meta system is seeing the conversion rates change during these two periods and using that information to push/pull spend based on changes in the expected value.
Anyone who has seen spend ramp unexpectedly during a sale moment or spend ramp down during a conversion tag issue has seen this play out in real time.
That’s the change in expected value dynamics at play.
Why the Meta system cheats and targets bottom of funnel users
The system will cheat if you let it.
You may notice a high percentage spend is going to ads that are at bottom of the funnel. Those heavy offer / conversion oriented ads.
IE: static image ads that say something like “Get $100 off”
This is because the Meta system is designed to help advertisers achieve whatever goal they set in the campaign settings, which is usually a purchase event.
And, the audiences most likely to complete the purchase event are typically retargeting audiences who respond well to ads that are heavy offer based. IE: “Get $100 off”.
In Meta’s eyes, these ads have a high expected value.
They’ve been to our website before, they (should) already know our story, and they may just need a nudge (offer) to complete the purchase.
In essence, we tell it to optimize for purchases, and left to its own devices, Meta goes after the lowest funnel audiences it can because it’s predicting this audience is the most likely to convert.
This is a problem for brand growth.
2 part system to drive brand growth through Meta
The first is through proper media buying.
This is primarily driven by how we are setting up our ad account in a way that gives the more upper funnel ads (less retargeting heavy) a chance to scale and reach new audiences.
The second is through making better ads (above) that are diverse enough to speak to different target personas, pain points and buyer stages.
It’s usually better to take bigger swings in terms of different ad concepts vs iterating on existing concepts.
This helps Meta reach net new people that we may not already be hitting with our current ad setup.
How does AI tooling fit into this
My POV on AI tooling right now is that it is a tool to help us create version one (v1) concepts and get them to market faster than we may have been able to in the past, but it is not a replacement of existing teams.
AI is an AND to our existing processes and teams - not an OR.
We still need humans to guide the AI tooling through deep understanding of the brand story, customer pain points, target messaging, and artistic style that continues to elevate the brand.
AI doesn’t understand the why behind the ads that work. It’s derivative.
In summary, AI should help us move quicker on ideating/concepting, but we still need to layer in judgement and taste to take something from v1 concept (AI enabled) to production ready (human driven).
Summary
In summary, Meta ranks your ads based on how likely they are to help you hit your goals - mainly using your bid, ad quality, and how likely people are to convert.
Left unchecked, it’ll chase easy wins at the bottom of the funnel (like retargeting offer ads).
If you want real brand growth, you’ve gotta structure your account to give upper-funnel creative a shot, and make ads that actually resonate.
AI can help speed things up, but the magic still comes from human taste.
I hope you’ve enjoyed this post!
Here are those resources I mentioned that worth checking out for further exploration…
Supporting resources:
AI Innovations in Meta's Ad Ranking Driving Advertiser Performance
This post is beyond the scope of this article but a must read (IMO)
This podcast is a helpful supporting piece to the above
Understanding conversion optimization in digital advertising