Open any ads dashboard and four acronyms greet you before your coffee's cold: CPM, CPC, CTR, CPA. They look like four separate scores. They're really one story told in stages, money turning into attention, attention into clicks, clicks into customers, and the leaders who read them well are the ones who know which number is a diagnostic and which is the one that actually pays the bills.

The quick version

  • CPM (cost per thousand impressions) is what you pay to be seen a thousand times, the price of attention. CTR (click-through rate) is the share of those views that turn into a click, clicks ÷ impressions.
  • CPC (cost per click) is what each click costs you; it falls out of CPM and CTR together. CPA (cost per acquisition) is what each customer or action cost, the number closest to the money.
  • They form a chain: cheaper attention or a higher CTR lowers your CPC; a better CPC or a higher conversion rate lowers your CPA. Fix the weak link, not the loudest number.
  • CPA looks like a result but is reported by an attribution model that often takes credit for sales you'd have made anyway. Trust it least, and check it against incrementality.

The idea in depth: four metrics, one funnel

These aren't competing metrics; they're four readings off the same pipe at different points. Start at the top. CPM, cost per mille, Latin for thousand, is, in Google's own definition, "a way to bid where you pay per one thousand views (impressions)." You're buying eyeballs in bulk. The formula is just cost ÷ impressions × 1,000: spend £500 to be shown 100,000 times and your CPM is £5. CPM is the price of raw attention, and it's the currency of brand and awareness campaigns where the goal is to be seen, not clicked.

Next, CTR, click-through rate. Google Ads puts it simply: the clicks your ad receives divided by the times it's shown, clicks ÷ impressions. Five clicks on 100 impressions is a 5% CTR. CTR is the only one of the four that isn't about money at all; it's about relevance. A high CTR says the right people saw something they wanted to act on. Google treats it as a quality signal in its own auction, it feeds the "expected CTR" component of Ad Rank, so a better CTR can literally lower the price you pay.

So read CPM and CTR together, before you touch anything else. A high CPM with a healthy CTR means attention is expensive but landing, a targeting or auction problem. A low CPM with a dismal CTR means you're cheap to show and nobody cares, a creative or audience problem. The two numbers diagnose different illnesses. Treat one as the other and you've found a fast way to waste a budget.

Now the two combine. CPC, cost per click, is what you're charged each time someone clicks. Google defines average CPC as "the total cost of your clicks divided by the total number of clicks." But notice what determines it: if you pay per impression, then CPC = CPM ÷ (1,000 × CTR). Your cost per click isn't an independent dial you set; it's the arithmetic consequence of how much attention costs and how often that attention converts to a click. Double your CTR and, all else equal, you halve your CPC. That's why "lower our CPC" is rarely a direct lever, it's an outcome of fixing CPM or CTR upstream.

"They are health metrics, not performance KPIs… inputs, not outcomes.", Search Engine Journal, on where CTR, CPC and CPM sit in the PPC measurement stack

Finally, the number leaders actually care about: CPA, cost per acquisition (or cost per action). Google's definition is "the total cost of conversions divided by the total number of conversions", what it cost you to get one purchase, signup or qualified lead. CPA is CPC divided by your conversion rate: if clicks cost £1 and one in twenty clicks converts, your CPA is £20. This is the metric closest to the money, because a "customer" is a thing the business can value. Set it against the gross margin or lifetime value of that customer and you finally have the only judgement that matters, are you acquiring people for less than they're worth?

Which is why a falling CPC is nothing to celebrate while CPA sits flat or climbs. Cheaper clicks that convert worse aren't progress; they're you buying more of the wrong traffic. Read the chain downward to CPA, and judge CPA against what a customer is worth, not against last month's CPA in isolation.

flowchart TD
  Spend("Ad spend") --> Imp(["Impressions, priced by CPM (cost per 1,000 views)"])
  Imp --> CTR(["CTR = clicks ÷ impressions: did the attention land?"])
  CTR --> Clicks(["Clicks, each one priced by CPC"])
  Clicks --> Conv(["Conversion rate: clicks that become customers"])
  Conv --> CPA(["CPA = cost ÷ conversions: what a customer cost"])
  CPA --> Worth("Compare to margin / lifetime value: was it worth it?")
The same money read at four points down the funnel, each metric a different stage, CPA the one nearest the business. Leaders Loop

Why the metric you trust most can mislead you most

Here's the uncomfortable part. CPM and CTR are nearly mechanical, an impression is counted, a click is counted, the arithmetic follows. (Even "impression" hides a wrinkle: the IAB and Media Rating Council's viewability standard only counts a display ad as viewable when at least 50% of its pixels are on screen for one continuous second, so "shown" and "seen" aren't the same thing.) But CPA, the number you'd most want to trust, depends on a judgement call no formula resolves: which ad gets the credit for a sale.

That judgement is the attribution model, and the common default, last-click, hands the entire credit to the final ad someone touched before buying. The problem, as practitioners have argued for years, is that last-click "credits conversions that would have happened anyway." Retargeting is the classic offender: showing an ad to someone already heading to checkout produces a flattering CPA and very little actual lift. One widely cited illustration: a campaign optimised straight to a £50 CPA hit the target comfortably while reducing revenue, whereas an incrementality-led approach missed the CPA target but drove real, net-new sales. The reported CPA looked worse and the business did better.

So treat CPA as a claim, not a fact, and test it. Incrementality asks the right question, would this conversion have happened if we hadn't run the ad? You answer it with a holdout: withhold the ads from a comparable group (often a set of geographies) and measure the difference in actual sales, not in attributed conversions. Most teams find some channels, brand search, retargeting, matter less than their attributed CPA suggests, and a few are doing more than they get credit for.

This connects to a deeper split in the evidence. Les Binet and Peter Field's effectiveness research (the Media in Focus work, building on The Long and the Short of It) argues that short-term metrics like CPA capture only half of advertising's job, the sales you convert now, and undervalue brand-building, whose payoff shows up later and rarely as a click. Optimise everything to this quarter's CPA and you quietly defund the work that makes next year's CPA cheaper.

The honest limitation runs the other way too: incrementality testing is harder, slower and noisier than reading a CPA off a screen, and small advertisers often can't run a clean holdout at all. None of this means CPM, CPC, CTR and CPA are useless, they're indispensable for steering a campaign day to day. It means they're navigation instruments, not the destination. The destination is profitable, incremental customers, and no single dashboard number proves you reached it.

A worked example

Take a direct-to-consumer tea brand, a composite, and every figure below is illustrative, running two campaigns with the same £2,000 budget. The marketing lead wants to know which is "winning."

Campaign A shows 500,000 impressions at a £4 CPM. It pulls a 0.5% CTR, so 2,500 clicks at a £0.80 CPC. Of those, 4% convert, giving 100 customers at a £20 CPA. Campaign B shows fewer impressions, 200,000 at a higher £10 CPM, because it targets a narrower, more expensive audience. But its creative is sharper: a 2% CTR, so 4,000 clicks at a £0.50 CPC. It converts at 5%, giving 200 customers at a £10 CPA.

Read it down the chain and the story tells itself. Campaign B pays more for attention (£10 vs £4 CPM) yet wins on every downstream number, because a four-times-better CTR more than pays for the pricier impressions, dragging CPC below A's despite the dearer start, and a higher conversion rate compounds the advantage into half the CPA. The naive instinct ("B's CPM is too high, kill it") points exactly the wrong way. The chain is what matters, not any single rung.

Now the twist that earns its keep. Suppose Campaign B is mostly retargeting people who already had tea in their basket. Run a holdout, withhold B from half your regions, and you discover those customers were going to buy anyway. B's true incremental CPA is £35, worse than A's. The dashboard crowned the wrong winner; the experiment corrected it. Same data, opposite decision.

flowchart LR
  subgraph A["Campaign A"]
    A1("£4 CPM, 500k views") --> A2("0.5% CTR → 2,500 clicks")
    A2 --> A3("£0.80 CPC")
    A3 --> A4(["4% convert → £20 CPA"])
  end
  subgraph B["Campaign B"]
    B1("£10 CPM, 200k views") --> B2("2% CTR → 4,000 clicks")
    B2 --> B3("£0.50 CPC")
    B3 --> B4(["5% convert → £10 CPA… but £35 incremental"])
  end
Illustrative: B looks twice as efficient on attributed CPA, until a holdout reveals most of its conversions weren't incremental. Leaders Loop

The lesson a leader takes from this isn't a number, it's a habit. Read every campaign as a chain from CPM to CPA, fix the weakest link rather than the noisiest one, and never let an attributed CPA make a budget decision a holdout could overturn. That habit is also where this topic plugs into the wider system: who you target sets your CPM and CTR before any creative is made (the work of segmentation, targeting and positioning), and what happens after the click, the offer, the price, the landing experience, is what your conversion rate, and therefore your CPA, really measures.

Frequently asked questions

What's the difference between CPM, CPC and CPA?

They price three different things. CPM is the cost of 1,000 impressions, you pay to be seen. CPC is the cost of a single click, you pay for interest. CPA is the cost of one conversion, you pay (in effect) for a customer or action. They also reflect who carries the risk: with CPM the advertiser bears the risk that nobody clicks; with CPA the risk of non-conversion sits more with the campaign's own funnel. Awareness campaigns tend to live on CPM, performance campaigns on CPC and CPA.

Is a high CTR always good?

Usually it signals relevance, the right people saw something worth acting on, and Google rewards it in the auction through expected CTR, which can lower your costs. But CTR can be high and worthless if the clicks don't convert: clickbait creative, a misleading headline, or accidental mobile taps all inflate CTR while doing nothing for CPA. Read CTR as a diagnostic, then always check whether those clicks survive to become conversions.

How do these metrics actually connect?

As a chain. When you pay per impression, CPC = CPM ÷ (1,000 × CTR), and CPA = CPC ÷ conversion rate. So a higher CTR lowers your CPC, and a better conversion rate lowers your CPA, without you ever bidding differently. That's why the useful question is rarely "how do I lower CPC?" but "which link in the chain, attention, relevance or conversion, is the weak one?"

Why shouldn't I just optimise everything to the lowest CPA?

Two reasons. First, attribution: a reported CPA often takes credit for conversions that would have happened anyway, so the lowest attributed CPA can be the least incremental spend, retargeting is the usual culprit. Second, timing: CPA only captures the sale you close now. Binet and Field's research shows brand-building pays off later and rarely registers as a click, so optimising solely to short-term CPA tends to defund the very work that makes future acquisition cheaper.

What is incrementality and do I need it?

Incrementality is the share of conversions that genuinely wouldn't have happened without the ad, the real lift, as opposed to the credit an attribution model assigns. You measure it with a holdout: withhold the ads from a comparable group and compare actual sales. You don't need it to run a campaign day to day, but you do need it before you make a big budget decision on the strength of an attributed CPA, because that number can point you confidently in the wrong direction.

Related in the Toolkit

Where to go next