Every paid channel sells attention by auction
There is no list price for a search click, a feed impression or a display slot on the open web. Each time a person searches, opens a feed or loads a page, the platform holds a short auction among the advertisers who are eligible to appear, and the winner pays a price set by that auction. The fundamentals hub describes the machine from the buyer's side: a budget buys entries into an auction or a placement, entries become impressions, impressions become clicks and clicks become results. This page opens the first step of that machine.
The reason to understand it is practical. The performance economics guide follows spend through to margin and sets the targets a programme must meet, but it treats the price of attention as an input. When that price moves, the cause can be the market, the season, the account's own creative or a setting nobody noticed, and each cause has a different fix. A buyer who runs search, social and programmatic together needs to read a cost change through the auction that produced it, rather than through a single idea of "the auction" borrowed from whichever platform they learned first.
The three auctions below are the ones the site's platform roots document in detail. Other platforms run their own: the LinkedIn root notes that auctions set what an impression or click costs there, and the TikTok root that auction ads are priced by competition. Their mechanics are not covered here.
Search auctions: bid, quality and Ad Rank
On Google search, every eligible ad receives an Ad RankThe value Google Ads calculates at each auction to decide if and where an ad shows. for each search, and Ad Rank decides whether the ad shows, where it shows and which assets show with it. Google documents the inputs, as the Quality Score guide sets out: the bid; the auction-time quality of the ad, built from expected click-through, ad relevance and landing page experience assessed for that particular search; the Ad Rank thresholds an ad must clear to show at all; the competitiveness of the auction; the context of the search; and the expected impact of the ad's assets and formats. The Ad Rank entry in the glossary keeps the same list for reference.
The price of a click (CPCAverage cost of each click.) comes from the same auction. Google describes it as what is needed to beat the Ad Rank of the next competitor below, adjusted for the ad's own quality, which is often less than the maximum bid. Quality therefore works twice: a stronger ad can rank above a higher bid, and it pays less for the position it reaches. The Quality ScoreGoogle's 1–10 keyword diagnostic. Google reports for each keyword is a diagnostic summary of those quality factors, not an auction input, which is why the guide treats it as a map of where quality is weak.
Two things follow for a buyer. A rising cost per click is either a real rise in the price of the same click or a change in which clicks are being bought, and the rising CPC diagnosis separates the two before blaming anyone. And competition is partly visible: auction insights lists the other advertisers that appeared in the same auctions, how often they ranked above you, and their share of impressions, though never their bids or budgets.
Feed auctions: bid, predicted action and ad quality
Meta's auction runs every time someone scrolls a feed, opens Stories or watches Reels, among the advertisers whose ad sets include that person. The Meta Ads pillar, drawing on Meta's documentation, states the rule: the ad with the highest total value wins, and total value combines the advertiser's bid, the estimated action rate (how likely the system believes this person is to take the action the ad set is optimised for) and ad quality, judged from people's feedback and from low-quality attributes such as clickbait. The highest bid does not automatically win.
Meta charges mostly per impression, whatever the ad set is optimised for, so the cost of a thousand impressions (CPMCost per 1,000 impressions.) is the price of entry. The what drives ad costs guide draws the consequence that matters most: the response to the ad is part of its price. An ad that earns a strong response competes well; a bland or tired one loses auctions on predicted action and quality rather than on bid, and its cost per thousand drifts up. The same guide notes that the bid is usually implied rather than typed. Under the default strategy the system bids to spend the budget on as many results as it can find, and a cost per result goal or bid cap set below what the auction requires reduces delivery rather than the price.
The difference from search is what is being auctioned. On search the advertiser bids on a query someone typed. In a feed nobody has asked for anything, so the auction prices a person, and the system's prediction of how that person will respond carries much of the weight a typed query carries on search.
Programmatic auctions: bid requests and first-price buying
On the open web, the auction is run by machines on both sides. The real-time bidding guide follows the sequence. The publisher's supply-side platform sends a bid request, a structured message defined by the OpenRTB specification that IAB Tech Lab publishes, describing one impression: the format and size, the site or app, the device, any user identifier and consent signal that is present, the floor price, and the time within which a response must arrive. Each demand-side platform holding a matching line item values the impression, for a performance line item by estimating the probability of the outcome it optimises towards and the value of that outcome, and returns a bid or declines. The exchange ranks the bids that arrived in time and above the floor.
Most major exchanges have moved to first-price auctionsAn auction in which the winner pays the amount it bid., where the winner pays what it bid; the request declares the auction type. Bidding the full valuation would then mean paying the full valuation every time, so most demand-side platforms apply bid shading: they estimate from their own history the lowest price likely to win and bid between that and the valuation. Not every impression goes to the open auction at all. The deal types guide explains that programmatic guaranteed deals take reserved inventory before any auction runs, preferred dealsA programmatic deal giving one buyer a first look at eligible impressions at a fixed, agreed price before the auction, with no volume guarantee. give a first look at a fixed price, and private auctions apply their own floors to invited buyers.
The programmatic buyer therefore controls more of the bidding machinery than a search or social buyer does, and carries more of the risk that goes with it. How a line item values an impression, whether it shades its bids and which deals it holds are all choices in the buyer's platform; the buying models guide covers the pricing models that sit on top.
Step by step: what the three have in common
Set side by side, the three auctions follow the same sequence, even though each names the steps differently.
- An opportunity appears. A search, an open feed or a page load creates one chance to show an ad.
- Eligible advertisers are found. Keywords, audiences, line item targeting and deal terms decide who may compete for it.
- A bid is set, usually by automation. Smart Bidding on Google, Meta's bid strategies and a demand-side platform's valuation each set the bid at auction time, from a target or the budget and from the conversion events they are fed.
- The bid is weighed or valued. Search adds auction-time quality; Meta adds predicted action and ad quality; a demand-side platform builds predicted response into its own valuation.
- A winner is chosen against the competition present for that opportunity.
- A price is set. Search charges what is needed to beat the next Ad Rank, adjusted for quality; Meta charges mostly per impression at whatever the auctions settle at; a first-price exchange charges the shaded bid.
| Row label | Search (Google) | Feeds (Meta) | Programmatic (open web) |
|---|---|---|---|
| What is auctioned | A search query at a moment | A person in a feed at a moment | One impression on a page or app |
| Who sets the bid | Smart Bidding from a target, or a manual bid | The bid strategy from a goal or cap, or implied by budget | The buyer's demand-side platform from its valuation |
| What adjusts the bid | Auction-time quality, thresholds, context, assets | Estimated action rate and ad quality | Built into the valuation; floors and deal terms on the sell side |
| How the price is set | What beats the next Ad Rank, adjusted for quality | Mostly per impression, at what auctions settle | First price, usually shaded |
| Where competition shows | Auction insights | Rising cost per thousand against your own history | Win rate and win prices in the platform |
Three common threads matter more than the differences. First, the advertiser rarely types a bid any more. The bid is set from a target and the events the system learns from, which is why the conversion events guide treats the event as a purchasing instruction. Second, relevance and response lower the price of a result on every channel: a better ad pays less per click on search, wins more cheaply in a feed, and makes each impression a demand-side platform buys more likely to produce the outcome it was valued on. Third, competition and season raise it. The cross-channel planning guide notes that cost per click and price per thousand typically climb around peaks as competitors arrive, which is why demand creation is bought in the weeks before.
What it means for you: what a buyer controls and what it does not
A buyer controls the inputs, not the market. The KPI tree framework is a useful way to see where each sits in the chain from price to result.
- The target. It sets the bid on every automated strategy. A target set too tight reduces delivery rather than buying the same results more cheaply.
- The events and values. They decide what the automation thinks a result is worth, and therefore what it bids for.
- The creative. It drives predicted response in feeds and expected click-through on search, and so the price paid for the same attention.
- The landing page. It is part of search quality and of the conversion rate on every channel; the message match guide covers keeping the click and the page in agreement.
- The breadth of audience or inventory. Narrow targeting gives the auction fewer opportunities to choose from.
- The budget. It decides how many auctions the campaign can enter.
What a buyer does not control is the other side: how many competitors want the same searches, people or impressions, and what those are worth to them. That is why every cost figure is read against the account's own history for the same period rather than against a published average, and why the optimisation hub judges a change on cost per result rather than on the price of attention alone. The CPC calculator and CPM calculator show what a change in price does to the volume a budget buys.
A soft drinks brand reads its costs across three auctions (illustrative)
Fizzlark Soft Drinks reads a pre-summer cost rise
Fizzlark Soft Drinks sells sparkling botanical drinks online and through a few retailers. In the six weeks before summer, its search cost per click rises from £0.62 to £0.81, its Meta cost per thousand impressions from £7.40 to £10.90, and its programmatic prospecting line's cost per thousand from £3.10 to £4.60. The first reaction in the weekly meeting is to raise every bid and every budget to hold volume.
The marketing lead reads each rise through its own auction instead. On search, the same keywords on the same devices cost more, and auction insights shows two new drinks brands ranking above Fizzlark on a third of its auctions; quality components are unchanged. That is competition, and the team decides to hold its target and defend only the brand and best-converting generic terms. On Meta, cost per thousand on the new-audience ad set has risen far more than on its other ad sets, and its one video has run for eleven weeks while link click-through has fallen from 1.4% to 0.8%. The ad is losing on predicted response, so the team launches three new hooks rather than raising the bid cap. On programmatic, the line item's shading had been switched off during a test and never restored, so it was paying its full valuation in every first-price auction it won. Shading is switched back on.
Four weeks later, search cost per click is still £0.79, which the team accepts as the season's price; Meta cost per thousand is back to £8.20 with the new creative; and the programmatic line's cost per thousand is £3.40 for a similar number of impressions.
Common mistakes
- Treating every auction as a search auction. A feed auction prices people and their predicted response, and a programmatic auction prices one impression at a time; reading them all as keyword bids leads to the wrong fix.
- Raising bids to fix a creative problem. An ad that has stopped earning a response loses on predicted action and quality; a higher bid buys the same weak response at a higher price.
- Setting a target below what the auction requires and calling it efficiency. On automated strategies a target or cap below the market price reduces delivery rather than lowering the price.
- Blaming the market without checking the mix. An average cost can rise because the account is buying a dearer mix of queries, people or placements, not because any single price rose.
- Ignoring the season. Competition rises predictably around peaks; compare with the same period last year before calling a rise a fault.
- Forgetting the programmatic settings. Shading, floors and deal priority are buyer-side and sell-side choices that move the price as much as competition does.
- Judging on the price of attention alone. A dearer click or impression that converts better can lower the cost per result; read both together.
