How Meta Ads actually works: the auction and delivery system
This series builds real, working Meta campaigns across 22 parts, using the same two businesses from this site's Google Ads series: Bright Leaf Coffee, an e-commerce coffee subscription business, and GreenDesk, a B2B project-management SaaS tool. Meta Ads shares almost nothing mechanically with search advertising — there's no keyword to bid on, no search intent to match. Part 1 covers the actual mechanism this series builds on: how Meta decides which ad to show whom, and why.
No search, no keywords — delivery instead of matching
Google Ads matches an ad to an active search. Meta Ads has no equivalent moment — there's no query to match against. Instead, Meta's system continuously decides, out of every ad currently running, which one to insert into a specific person's feed at a specific moment, based on a prediction of how that specific person is likely to respond to that specific ad.
The auction: still an auction, different inputs
Total Value = Advertiser Bid × Estimated Action Rate × Ad QualityStructurally similar to Google's Ad Rank from its own series — bid isn't the only factor — but "Estimated Action Rate" replaces search relevance entirely: Meta's system predicts, for this specific person, how likely they are to take the action the campaign is optimized for (a click, a purchase, a lead form completion), based on that person's real behavior history and the ad's own performance so far with similar people.
What "Ad Quality" actually measures
Ad Quality draws on real, direct signals: how people react to the ad (hiding it, reporting it, engaging positively), and how the destination — a website or Instant Form — actually performs once someone gets there. This is Meta's rough analog to Google's Quality Score (covered in the Google Ads series), and it carries the same real, practical consequence: a low-quality, poorly-received ad costs more to deliver the same result than a well-received one, even at an identical bid.
A concrete example: two ads targeting the same audience
Bright Leaf Coffee Ad A: bid $8 CPM, high positive engagement, low hide rate
Bright Leaf Coffee Ad B: bid $12 CPM, mediocre engagement, some "hide this ad" clicksAd A can out-deliver Ad B despite a meaningfully lower bid — Meta's system is actively trying to show people ads they're predicted to respond well to, and a pattern of people hiding or scrolling straight past an ad is itself a real, negative signal factored directly into future delivery, not just an abstract quality label.
The account structure this series builds toward
Business Manager (Bright Leaf Coffee)
└── Campaign (Sales — Coffee Subscriptions)
└── Ad Set (Core Audience — Coffee Enthusiasts)
└── Ad (Carousel — Subscription Plans)
└── Ad Set (Retargeting — Cart Abandoners)
└── Ad (Single Image — Complete Your Order)Campaign, then Ad Set, then Ad — Meta's three-level hierarchy, covered fully in part 4. Audience targeting and budget live at the Ad Set level (a genuinely different placement than Google's keyword-per-ad-group structure), and creative lives at the Ad level — a structural difference that shapes almost every decision in the rest of this series.
Assuming Meta Ads works like a smaller, simpler version of Google Ads with the same underlying logic. The two platforms solve fundamentally different problems — matching intent versus predicting response — and techniques that work well on one (keyword match types, for instance) have no direct equivalent here at all.
Next: setting up Meta Business Manager and Bright Leaf Coffee's actual first campaign, from a blank account to a live Sales campaign.