Ravelin Sails repairs and cuts sails for small boats out of a loft in Plymouth, and it bids on the phrase mainsail repair against a national chandlery with a great deal more money. It bids £3.20. The chandlery bids £5.50. Ravelin wins the auction and pays £2.76.
That is not a quirk and it is not luck. It is the rule the whole of this topic runs on, and almost nobody buying clicks for the first time knows it, which is why the first instinct when an ad stops showing is to raise the bid.
The auction, worked
Every advertiser in the auction gets a rank. The published version of the rule is the bid multiplied by a quality figure the platform assigns, and the real one has more terms in it than Google names publicly. The arithmetic survives the extra terms, because what it demonstrates is not the exact number but the ordering: the winner is the highest bid times quality, not the highest bid.
Ravelin bids £3.20 and carries a quality of 8, which is 25.6. The chandlery bids £5.50 at a quality of 4, which is 22.0. A marketplace listing bids £2.90 at 7, which is 20.3, and a rival loft bids £2.10 at 6, which is 12.6. Ravelin sits top with the second cheapest bid on the board.
Then the price. What you pay is the rank below you divided by your own quality, plus a penny. The rank below Ravelin is 22.0 and Ravelin's quality is 8, so the click costs £2.76 rather than the £3.20 offered. A £3.20 bid with quality 8 paid £2.76, and the missing 44p was never available to be spent: the chandlery's weak ad set the price, and a competitor being bad at this makes your clicks cheaper.
Quality is a price, not a grade
The quality figure looks like a report card and behaves like a discount. Push Ravelin's 8 to a 9 and the same auction, against the same bidders, charges 22.0 divided by 9 plus a penny. Raising quality from 8 to 9 drops the same click to £2.45. Nothing about the bid changed, nothing about the competition changed, and the loft is now buying the same visit for 31p less.
What goes into that figure is mostly two things a small advertiser controls completely: whether the ad says what the searcher asked for, and what happens on the page they land on. The landing page is inside the quality score and a slow one raises every click, which makes where the click arrives, and usually leaves a bidding decision rather than a design one. This is the single most expensive thing that a beginner does not know, and the full version is at bid, quality and what you actually pay.
The two lists nobody reconciles
You bid on keywords and you are charged for search terms, and they are different lists. The keyword is what the account was told to target. The search term is what somebody actually typed, and the platform decides how loosely those two are allowed to match. Ravelin bid on mainsail repair and paid, over six weeks in 2026, for mainsail repair cost, which is fine, and for how to repair a mainsail yourself, which is somebody who has already decided not to buy anything.
The search terms report is where the second list lives and it is the first screen to open in a new account. Everything else in telling the platform what a result is worth is downstream of having stopped paying for searches that were never yours.
The order, which is set by what stops money leaving fastest
Search rewards patience and this does not. The meter runs while you learn, so the order here is not a dependency, it is a triage: whatever stops the bleeding soonest, first. That puts the search terms report ahead of the landing page, the landing page ahead of the writing, and the bid, the thing everybody reaches for on day one, last of the four.
The reason the bid comes last is arithmetic rather than taste. There is usually a few pence of room in it, and there is often a third of the spend in the search terms report. Ravelin's first month bought 214 clicks, of which 61 were people looking for instructions, and no bid adjustment in the world recovers that £168.
The order
Four rungs, ordered by how fast each one stops money leaving
Not a dependency and not a preference. The meter runs while the account is being learned, so the order is triage.
- 04Then argue about the bidThe lever reached for on day one, holding the least room of the four.Last
- 03Rewrite the ad to match the pageCheap to do, slow to read, and the only rung with a creative argument on it.Month one
- 02Fix the page the click lands onIt sits inside the quality figure, so it moves the price and the result together.Week two
- 01Stop paying for the wrong searchesThe search terms report, read once. It is a different list from the keywords.Week one
The rungs are in the order that pays back soonest, which is not the order of importance. A bid set badly enough will lose money on all four.
Where this stalls: the account that is optimised and unprofitable
The stall in paid advertising is not incompetence and it usually arrives after several months of visible improvement. The cost per click comes down, the click-through rate goes up, the quality figures go green, and every one of those numbers describes the auction rather than the business. An account can be excellent at buying clicks for a product that cannot afford them, and it will report as excellent right up to the point somebody totals the orders.
The vanity metric here is cost per click on its own. It is the number the platform leads with, the number an agency reports monthly, and the only one in the account that can be improved by buying worse traffic more cheaply. What it has to be read against is what a customer is worth, which is not in this topic at all and belongs to the decisions that come before any channel. Reading it properly is knowing what the money bought.
What to open in 2026
Google Ads is where the intent is, and it replaced Google AdWords, renamed in 2018, so anything still using the old name is older than the interface it is describing. Microsoft Advertising replaced Bing Ads, renamed in 2019, and it carries a smaller audience at a lower price. Meta Ads Manager replaced boosting a post from the page itself, which is the same auction with most of the controls taken away. None of the three has a free tier and saying otherwise would be a lie: the tools cost nothing and the clicks are the product.
The change worth knowing about is that since about 2023 the default recommendation inside these accounts has been the campaign types that hand targeting over to the platform entirely. They trade control for reach, they are very hard to diagnose when they fail, and they are the reason the search terms report matters more now than it did when every keyword was chosen by hand.
Ten clusters, six of them platforms
The auction and then the two clusters that set the price come first. After those, the platforms divide by what you can buy on them. Intent you can buy, at the moment it appears is the only one where the customer arrives already asking, which is why it is the expensive one. Facebook and Instagram, and interest based targeting sells to people who were not looking, and the ad has to do work that a search ad never has to do. Cheaper clicks, and an older audience is the one a small advertiser tries second and the one most often left switched off after a fortnight for no reason.
The remaining three are narrower. Expensive, precise and business to business only makes sense above a certain deal size, and ads that have to look like content is a production problem before it is a media one. Underneath all six sits the ad itself, which decides most of the result, because the ad that wins the auction is not always the ad that wins the click.
Sideways, the unpaid version of the same results page is being the result someone clicks, without paying for it, and the counting that makes any of it checkable is measuring it, so decisions are not guesses. This topic is one of the four under only four of them are channels.
The open question at Ravelin is whether the 8 can be moved at all. It has sat at 8 since the account opened, the page it points at was rebuilt twice, and neither rebuild shifted it. Nobody outside the platform can see which of the components is holding it down, which means the cheapest lever in the account is also the one nobody can aim at.