Three different numbers wear the name cost per click, and Google documents them on three separate pages. Your maximum CPC is a ceiling you enter. Your actual CPC is what you are charged for a given click. Your average CPC is a division performed after the fact.
Almost no page separates them. It is nevertheless the distinction that decides what you are entitled to conclude from any of the three.
This page covers the three objects, why the third is an amount nobody was ever charged, where the benchmarks in circulation actually come from, and what the two largest platforms document about their own pricing.
Max CPC, actual CPC, average CPC
They answer different questions, and stacking them under one word is how budget conversations go wrong.
The maximum CPC is a ceiling. It is the most you accept to pay for a click. Google’s documentation is explicit that you are often charged less, sometimes much less, than that figure.
The actual CPC is a transaction. It is what you were charged for one specific click, in one specific auction. It is the only one of the three that corresponds to money actually moving.
The average CPC is a statistic. Total cost of clicks divided by total number of clicks, computed after the period closed. It describes what happened, and it prices no future click.
Why the confusion is expensive. People set a maximum bid by reference to an average benchmark, which compares a ceiling to a statistic. Or they read an average CPC as the price of the next click, which it never is. Both errors lead to the same behaviour: setting bids too low out of caution, and losing auctions rather than saving money.
The average CPC is an amount nobody was charged
The demonstration is not ours. It is the worked example in Google’s own documentation.
The definition. Average cost-per-click “is calculated by dividing the total cost of your clicks by the total number of clicks”, per Google’s documentation.
The example Google supplies itself. “Let’s say your ad gets two clicks, one costing $0.20 and one costing $0.40, for a total cost is $0.60. You’d divide $0.60 by 2 to get an average CPC of $0.30.”
What that establishes. Neither of the two clicks cost $0.30. The average CPC is an output statistic, not a tariff, and it corresponds to no transaction that took place.
Why this is more than a curiosity. A sector CPC benchmark is an average of these averages. It aggregates accounts each of which already published a number matching nothing that was charged. The gap between that figure and what you will pay is not imprecision, it is a change of nature.
The practical consequence. An average CPC is useful for tracking a trend on your own account, month to month, at constant structure. It is not useful for predicting what your next click will cost, nor for comparing yourself to anyone.
Your “average” benchmark is a median, and the source says so
This is the finding that should change how the figure is quoted, and it takes one paragraph to verify.
What the most republished US benchmark declares. WordStream publishes its methodology in a block on the page: “This report is based on a sample of 13,474 US-based search advertising campaigns running between Apr 1, 2025, and March 31, 2026. Each subcategory includes at minimum 52 unique active campaigns. ‘Averages’ are technically median figures to account for outliers. All currency values are posted in USD.”
Read that middle sentence again. The publisher states plainly that the numbers labelled averages are medians. Using a median to resist outliers is good practice. Republishing it as an average is not.
Why nobody downstream mentions it. The same figures, $5.42 across industries, $9.87 for legal services, $8.00 for dentists, $3.39 for finance and insurance, are republished on a more accessible page that carries no methodology block at all. No sample size, no period, no median disclosure. The version that is easiest to cite is the version stripped of what makes it citable.
What is lost in that step, concretely. The 13,474 campaign sample. The April 2025 to March 2026 window. The US-only perimeter. The minimum of 52 campaigns per subcategory. And the statement that these are medians.
Why the median point matters beyond pedantry. A median and a mean answer different questions on a skewed distribution, and advertising costs are heavily skewed. Quoting a median as an average tells your reader that half the accounts sit above and half below, when they think they are being told what a typical account spends in total. Those are not the same claim.
The rule that follows. Before quoting any CPC figure, open the page it came from and look for a methodology block. If there is not one, the figure may still be correct, but you cannot say anything about what it describes.
Neither Google nor Microsoft publishes an average CPC
The two companies holding the data decline to publish it, and one of them explains why.
What Google writes about its own prices. “Prices are dynamic, which means they’ll change from instance to instance”, in its cost documentation. Google publishes no average cost per click anywhere: not globally, not by sector, not by country.
What Microsoft does. The same, which is to say nothing. Its documentation describes a maximum bid and states that actual cost may be lower depending on competition, without ever publishing a reference value.
Where the two converge, against what is widely taught. Google states that Quality Score “is not an input in the ad auction. It’s a diagnostic tool.” Microsoft is blunter still: “It is not used at auction time to determine ad rank, and does not affect cost or spend.” Two competitors documenting the same thing independently is about as solid as this kind of claim gets.
What that leaves. Every reference figure you read comes from a third party aggregating the accounts it manages. That is not disqualifying, and it is worth knowing before anyone treats such a figure as a target.
The question to ask of any CPC figure. Who aggregated what, over what period, and is it a mean or a median? If the page does not answer, the number is not a measurement you can use.
What actually decides what you pay
Four things, and your bid setting is only one of them.
The competition on your query. How many advertisers want the same search today. This moves with your market and with the calendar, and it is the largest single factor on most accounts.
Where you sit in the ranking. You pay the minimum required to clear the thresholds and beat the ad immediately below you. If nobody qualified sits below you, you pay the reserve price.
How your ad and page are received. The auction weights predicted engagement and quality alongside the bid, which is why a better ad can hold a better position at a lower cost than a competitor bidding more.
The two settings that break the ceiling. Enhanced CPC and bid adjustments can push actual cost above your stated maximum. The ceiling stops being a ceiling the moment you delegate part of the setting.
One term to stop using. You will read everywhere that Google Search runs a second-price auction. Google never writes it, and when it moved Ad Manager to first-price bidding in 2019, the announcement stated the change would have no impact on auctions for ads on Google Search, without saying what those auctions were instead.
Why a sector CPC table destroys the information you need
The tables are the most consulted part of the subject and the least usable, for a reason that is arithmetic rather than editorial.
The first aggregation. Inside one account, every click has its own price, and they are collapsed into one figure. Dispersion is lost at step one, which is what Google’s two-click example shows in miniature.
The second aggregation. Thousands of accounts, each already collapsed, are collapsed again into one sector figure. Whatever separated a well-targeted account from a badly targeted one has now disappeared twice.
The third aggregation, the one nobody mentions. Sector labels are broad. “Finance and insurance” holds a mortgage broker and a corporate treasury software vendor, whose auctions have nothing in common. The label suggests a peer group that does not exist.
What survives all three steps. An order of magnitude, and a direction of travel. That is genuinely useful, and it is much less than what the tables are used for.
What does not survive. Any statement about what you should be paying. The number cannot carry that claim, because everything that would make it specific to you was averaged away three times.
The honest use of a sector table. To check whether you are in a wildly different order of magnitude from the market, which occasionally reveals a targeting error worth investigating. Not to set a target, and never to judge a vendor.
What a cost per click never tells you
The metric stops at the click, and most of the questions people ask it are downstream of that point.
It says nothing about who clicked. A click from a student researching your industry and a click from a purchasing manager cost the same and are worth entirely different amounts.
It says nothing about what the page did. A cheap click that lands on a page answering a different question is a cost with no counterpart.
It says nothing about the deal. No revenue, no margin, no sales cycle enters the calculation, which is why a cost per click can improve for four straight months while the business gets worse.
Why that is not a criticism. A cost of traffic is a legitimate thing to know, and knowing it precisely is useful when you are buying traffic on purpose. The error is asking it to report on outcomes it was never built to see.
The pair that actually decides. Cost per click and cost per qualified enquiry, read on the same rows and the same period. When the first falls and the second rises, you bought cheaper traffic that was worth less, which is the single most common pattern in B2B search accounts.
The only cost per click that concerns you
Once you set aside what does not exist, little remains, and it is enough.
Yours, compared with itself. On your own account, at constant campaign structure, month to month. It is the only series that holds your sector, your geography, your keywords and your real competition constant.
Read next to a second number, never alone. The cost per closed deal. A click at $7 on a purchase-intent query beats a click at $0.50 on an out-of-scope audience, and only the second number tells you which you have.
What should alert you in a report. A falling cost per click presented as a result. It frequently comes from budget shifting toward broader, cheaper queries, which costs more per deal while displaying a better cost per click.
The question that settles it. Our cost per click fell: what changed in the queries we are buying? If nobody can answer, the fall is not good news, it is an unknown.
When the figure is genuinely useful. As a break detector. A cost per click that jumps within days, with no change on your side, tells you a competitor entered your auctions or the season turned. That is worth knowing, and it is the job the metric does well.
Three ways a cost per click falls without anything improving
All three work within days, and none of them buys you a better click.
Shift budget toward broader, cheaper queries. Generic head terms clear at a lower price than precise purchase-intent ones. Move spend that way and your cost per click improves while the share of clicks from people who could buy collapses.
Add a cheaper network or placement. Search partners, automatic placements, or a display extension attached to a search campaign all deliver clicks at a lower unit price. Blended into one report, they pull the average down and tell you nothing about which line produced the enquiries.
Let the mix drift toward brand terms. Clicks on your own brand name are cheap because almost nobody bids against you. If brand traffic grows as a share of the account, your blended cost per click falls without a single non-brand auction having changed.
What makes these hard to catch. None is a mistake, and two of them are sometimes exactly the right call. They simply move a ratio for reasons unrelated to how well you are buying.
The split that exposes all three. Report brand and non-brand separately, and report each network separately. Most accounts that look like they improved their cost per click improved their reporting mix instead.
The question to ask. Our cost per click fell: which queries and which networks did the extra clicks come from? If the answer is not immediate, nothing has been established.
How to read your own cost per click over time
Four columns, and most reports have two of them missing.
Cost per click, split by brand and non-brand. Blending the two produces a number that moves whenever your brand demand moves, which is not a statement about your buying.
The volume behind each line. A cost per click computed on forty clicks moves for reasons that have nothing to do with the market. Show the click count next to the price, always.
A dated change log. Every keyword change, every match type change, every bid or budget change, with the date. Without it, an explanation offered in a monthly review is a reconstruction after the fact rather than an attribution.
Cost per qualified enquiry, on the same rows. This is what turns a traffic price into a diagnostic. It is also the column that gets dropped first when reporting is built by whoever runs the campaigns.
The cadence. Monthly, compared against the same month a year earlier. Weekly readings on a B2B account measure noise, and month-over-month comparisons mostly measure the advertising calendar.
What that setup lets you say. Not that the cost per click is good or bad, which are not statements about anything. But that non-brand cost per click rose 20% year over year on stable volume while cost per qualified enquiry held, so the auctions got more expensive and the targeting absorbed it. That is a sentence someone can act on.
Where to go next
You want to structure the account this sits in. Google Ads account structure for B2B.
You are buying exposure rather than clicks. CPM, cost per thousand impressions.
Your click-through rate is moving and you do not know if it is real. Click-through rate benchmarks.
You want the cost of a contact, not a click. Cost per lead.
You want the whole chain and its vocabulary. Media buying explained.
You want to know which metric to run on. ROAS vs MER vs CAC vs LTV.
In short
- Three objects wear one name: the maximum CPC is a ceiling, the actual CPC is a transaction, the average CPC is a statistic computed after the fact.
- The average CPC is an amount nobody was charged. Google’s own example: $0.20 and $0.40 give $0.30.
- Your “average” benchmark is a median, and the publisher says so in a methodology block: 13,474 US campaigns, April 2025 to March 2026, “‘Averages’ are technically median figures.”
- The version everyone cites drops that block. Same numbers, no sample, no period, no median disclosure.
- Neither Google nor Microsoft publishes an average CPC, and Google writes that prices change from instance to instance.
- Quality Score is not an auction input, in Google’s words, and Microsoft states it does not affect cost or spend.
- Your maximum bid is a ceiling, not a price. Bidding low to be careful loses auctions rather than saving money.
- Google never calls Search a second-price auction. Its 2019 move to first-price bidding in Ad Manager explicitly excluded Search.
A cost per click is not judged against a benchmark. Book a diagnostic, or see how we approach B2B paid acquisition.