How ad auctions actually work

First, the money question — answered

You will not spend a euro in this module. Not in this lesson, not in the project. Every campaign you build stays in draft; the module project is a €500 plan you design, defend and never run. A time-poor person with no budget can complete all of it, and the plan they leave with is exactly what they would hand a budget-holder the day money appears.

That's not a compromise. In paid advertising, the plan is where the skill lives. Spending is the easy part — platforms have made it one button. Knowing what the numbers must be before you press it is the part that keeps you employed, and it is entirely learnable on paper.

The auction you never see

Every ad you have ever been shown was sold in an auction that ran in the milliseconds before the page loaded. Someone searches, or opens a feed. An ad slot exists for a fraction of a second. Every advertiser whose targeting says "I want this person" is entered, ranked, and one wins.

Two things about this auction matter more than everything else in the module.

First: it happens per impression. You are not buying a billboard for a month. You are bidding for one person's attention, one moment at a time, against whoever else wants that same person. This is why small budgets can genuinely compete — a €10-a-day advertiser bidding on exactly the right people is, in each individual auction, on equal footing with a national brand bidding lazily. Budget buys volume. It does not buy victory.

Second: the winner typically pays just enough to beat the next bidder, not their own maximum. Your bid is a ceiling, not a price. Platforms run variations of this second-price logic because it lets you bid what a click is honestly worth to you without being punished for honesty.

Why the highest bid does not win

Here is the part most beginners never learn, and it changes everything: the auction does not rank by bid. It ranks, roughly, by

bid × predicted quality

where "predicted quality" is the platform's estimate of how likely this ad is to get a response from this person. Google calls its version Ad Rank, driven by Quality Score. Meta calls it total value, driven by estimated action rates. The vocabulary differs; the shape does not.

Why do platforms do this? Because their product is their users' attention, and an ad nobody wants to see burns that product. A platform that auctioned slots to the highest bidder regardless of relevance would fill its feeds and results pages with junk, and its users would leave. So every major platform taxes boring ads and discounts relevant ones.

The consequence is the sentence this lesson exists to deliver: the ad is the product. You cannot buy your way past a weak ad, a wrong audience or a broken landing page — the auction reprices your weakness upwards every single impression. And you can out-place richer competitors with a better ad, because relevance is a discount.

One auction, slowed down

A Graz resident searches "physiotherapie graz". Three advertisers are entered. The quality numbers below are illustrative — platforms don't publish them per auction — but the mechanism is real.

                       bid      predicted quality   rank (bid × quality)
Herzog Physio         €1.50   ×        9          =      13.5
National directory    €4.00   ×        3          =      12.0
Voucher marketplace   €2.00   ×        2          =       4.0

Herzog wins. Tomas bid barely more than a third of the directory's €4.00, but his ad names the city and the service, and it lands on his Check-up booking page from Module 4 — not a generic homepage. The directory's ad is the same ad it shows in every city, pointing at a listings page. The system predicts, correctly, that the searcher wants Tomas.

And what does Tomas pay? Just enough to beat rank 12.0:

12.0 ÷ 9 ≈ €1.33  → he pays about €1.34, under his €1.50 ceiling

For the directory to overtake him at quality 3, it would need to bid above 13.5 ÷ 3 = €4.50 — and pay it, every click, forever. That is what "relevance is a discount" means in euros: Tomas's quality lets him occupy the top position at less than a third of the price his competitor would need. This is a simplified model — the real formulas add factors — but the shape is right, and the shape is what sets your costs.

Buying demand versus making demand

Paid channels split into two families, and they are different jobs with different maths.

Intent channels — search ads, mostly — harvest demand that already exists. Someone typed "physiotherapie graz"; the need is declared in their own words. Conversion rates are higher, volume is capped at however many people search, and the craft is in not wasting money on the wrong searches. Lesson 2.

Interruption channels — social, display, video — create or redirect demand. Nobody asked to see the ad; they were doing something else. Conversion rates are lower, volume is nearly unlimited, and the craft is in the creative. Lesson 3.

Most businesses need to know which family fits before choosing a platform. If people already search for what you sell (your Module 6 keyword research answers this), buy that intent first — it is the cheapest demand you will ever reach. If nobody searches for it because they don't know it exists, no search budget can help you; interruption is the only door.

Is your funnel ready to be amplified?

Paid traffic is an amplifier. It scales a working funnel — and it scales a broken funnel's losses, at whatever CPC the auction charges. Before any plan, run the readiness test: an offer that has survived contact with customers (Module 3), a landing page built for the conversion (Module 4), follow-up that catches the people who don't buy today (Module 8), and tracking that would show you — in data, not in feelings — if conversions doubled.

Lena at Storkflow fails one item, and her decision is the lesson's second teaching point. She already spends €400 a month on Google Ads, and since Module 4 those clicks at least land on a real page. But her trial follow-up automation went live only weeks ago and her source tracking is still partial. Her verdict: keep the existing spend running, refuse to scale it for a quarter, and let the funnel accumulate evidence first. Deciding not yet is a passing grade in this module. Amplifying an unmeasured funnel is the failure.

Your exercise now is the same assessment, for your business. No money required — just honesty.

---