The chain
Almost every campaign, on any platform, can be described by the same sequence of counts:
- Spend
- Impressions
- Clicks
- Conversions
- Revenue
- Profit
The counts are facts the platform records. The metrics are ratios between them. Once you see them that way, the formulas stop being things to memorise and become a way of asking "which link moved?"
Exposure metrics
Impressions — the number of times an ad or piece of content was shown. An impression does not mean it was seen; viewability metrics address that. Impressions are the denominator of several ratios and rarely a goal in themselves.
Reach — the number of distinct people (or devices, or browsers) who saw it at least once. Reach is what an awareness campaign is buying; impressions divided by reach gives frequency, the average number of times each person was exposed.
Frequency = Impressions ÷ Reach
Frequency is a control: too low and the message is not remembered, too high and the audience is fatigued while the budget is spent re-reaching the same people. The reach and frequency calculator shows how the two trade off at a fixed budget.
Response metrics
Click-through rate (CTRShare of impressions that result in a click.) — clicks as a share of impressions. It measures how well the creative and the targeting match: a low CTR says either the wrong people saw it or the message did not interest the right ones.
CTR = Clicks ÷ Impressions × 100(%)
CTR is only comparable within the same format and placement. Search ads, shown to people who asked for the thing, earn far higher click rates than display ads shown to people doing something else; comparing the two says nothing about either.
The CTR versus conversion rate comparison shows how to read the rates either side of the click together, so a weak result can be traced to the ad or the page.
Cost metrics
Cost per thousand impressions (CPMCost per 1,000 impressions.) — the price of exposure. It is set by auction: how many advertisers want the same audience, how good the platform predicts the ad will be, seasonality, and the placement.
CPM = Spend ÷ Impressions × 1,000
Cost per click (CPCAverage cost of each click.) — the price of a visit. In auction platforms it is the product of exposure price and response rate: an expensive audience with a strong creative can be cheaper per click than a cheap audience nobody clicks.
CPC = Spend ÷ Clicks = CPM ÷ (1,000 × CTR)
Cost per acquisition (CPACost of each conversion.) — the price of a conversion (a sale, a lead, a sign-up — whatever the campaign was set to achieve). It extends the same decomposition one step further, which is why a rising CPA can be diagnosed by asking which of the three underlying rates moved.
CPA = Spend ÷ Conversions = CPM ÷ (1,000 × CTR × CVR)
The CPA calculator lets you move each component and see the effect.
Outcome metrics
Conversions — completed actions the campaign was aiming for. The definition is a choice: a purchase, a qualified lead, an app install, a sign-up. A campaign's reported conversions mean nothing until you know which action was counted and over what window.
Conversion rate (CVRShare of visits or clicks that convert.) — conversions as a share of clicks or sessions. It measures the landing experience and the offer more than the ad: the ad got the person there, the page converted them or did not.
CVR = Conversions ÷ Clicks (or Sessions) × 100(%)
Average order value (AOVAverage revenue per order.) — revenue per order. Raising it (bundles, upsells, thresholds) often moves return more cheaply than raising conversion rate.
Return metrics
Return on ad spend (ROASRevenue generated per unit of ad spend.) — revenue attributed to a campaign divided by its spend. It is the standard way to compare campaigns, and it is incomplete: it ignores product cost, so a high ROAS on a low-margin product can still lose money.
ROAS = Attributed revenue ÷ Ad spend
Break-even ROAS (Break-even ROASMinimum ROAS to avoid a loss.) — the ROAS at which the campaign neither makes nor loses money, given the margin. It is the number ROAS should be compared with, and most teams do not know theirs.
Break-even ROAS = 1 ÷ Gross margin
Return on investment (ROIProfit relative to cost.) — profit relative to total cost, including product cost, fees, tools and people. It is the business's metric; ROAS is marketing's. The ROAS calculator computes both and the break-even point.
ROI = (Revenue − Total cost) ÷ Total cost × 100(%)
Customer acquisition cost (CACCost to acquire one new customer.) — all acquisition cost (media, agency, tools, people) divided by new customers in a period. CPA is per conversion on one campaign; CAC is per customer across the programme.
Customer lifetime value (LTVTotal value a customer brings over time.) — the margin a customer generates over the relationship. Compared with CAC it answers the question ROAS cannot: whether acquiring customers at this price is worth it once repeat purchases are counted. The lifecycle marketing guide covers the levers that raise it.
Reading the chain on a Meta campaign
A campaign's CPA has risen since last month. CPM is flat, so exposure did not get more expensive. CTR has fallen, so fewer of the people who saw the ad clicked — the creative has fatigued or the audience has shifted. CVR is unchanged, so the landing page is doing its job. The fix is creative, not the page and not the bid. Without the decomposition, the same report would have prompted a bid change that solved nothing.
Vanity metrics and decision metrics
A vanity metricMetrics that look impressive but do not inform a decision or track a business outcome. moves without telling you what to do. Impressions, followers, page views and "engagement" are the usual suspects: they can rise while the business stands still, and they are easy to inflate. A decision metric changes a decision when it changes: cost per acquisition changes bids and budgets, conversion rate changes pages, frequency changes audience size, break-even ROAS changes whether a campaign runs at all.
The test is not the metric but the question attached to it. Impressions are a vanity metric on a direct-response report and a decision metric on an awareness campaign whose job is reach. The objectives and KPIs guide covers how to attach the question.
Platform metrics and analytics metrics
The same name can mean different things in different tools. A "conversion" in an ad platform is counted by the platform's attribution rules — a window of days after a click or a view, sometimes modelled when tracking is incomplete — while a conversion in web analytics is an event that happened on the site. Clicks in a platform and sessions in analytics differ because of bots, accidental clicks, blocked tracking and page abandonment. Neither source is the truth; a measurement plan names which source is the reference for which metric so that reports do not argue with each other.
Inside analytics, the sessions versus users comparison explains why the same conversions give a different rate depending on which count sits underneath.
Quick reference
| Metric | Formula | Use it to | Watch out for |
|---|---|---|---|
| Impressions | count | denominator for rates | not the same as views |
| Reach | distinct people | awareness delivery | device and browser double-counting |
| Frequency | impressions ÷ reach | exposure control | fatigue above the useful range |
| CTR | clicks ÷ impressions | creative and targeting fit | only comparable within a format |
| CPM | spend ÷ impressions × 1,000 | exposure price | rises with competition and season |
| CPC | spend ÷ clicks | visit price | hides whether visits convert |
| CVR | conversions ÷ clicks or sessions | landing experience and offer | which conversion, and which denominator |
| CPA | spend ÷ conversions | efficiency per action | which window, which attribution |
| AOV | revenue ÷ orders | value per sale | mix changes masquerading as uplift |
| ROAS | revenue ÷ spend | compare campaigns | ignores margin; platform-attributed revenue |
| Break-even ROAS | 1 ÷ margin | the line ROAS must clear | margin varies by product |
| ROI | (revenue − cost) ÷ cost | whether the activity pays | needs full cost, not media only |
| CAC | acquisition cost ÷ new customers | programme-level efficiency | which costs are included |
| LTV | margin over the relationship | what a customer is worth | needs time to observe; use proxies early |
Two conventions keep the table usable in practice: state the denominator every time a rate is quoted (per click or per session), and state the window every time a conversion is counted. Half of the arguments about these numbers dissolve once both are written on the report.
Before comparing any of these metrics with a published figure, read the guide to using industry benchmarks.
When visits themselves are the goal, the website traffic objective hub shows how these response and cost metrics combine into cost per engaged session and which channels deliver visitors who stay.
When a partner's report climbs the platform links of this chain while the business outcome at its end stays still, the diagnosis of agency reports that look good while sales are flat works through the usual causes.
When visits climb and orders do not, the traffic up but sales flat diagnosis traces the gap to low-intent sources, bots, mismatched pages, a broken step, offline sales or a traffic target.
Where visits are the objective, the KPI tree for website traffic arranges these response and cost metrics into channel branches that add up to engaged visits, with cost per engaged visit as the guardrail.
Common mistakes
- Comparing rates across formats. CTR, CVR and CPM are only comparable within the same channel, format and stage of the funnel.
- Celebrating ROAS without margin. A campaign above break-even on one product can be below it on another at the same ROAS.
- Optimising the last link only. Lowering CPA by narrowing to brand search and retargeting looks efficient until the pool of ready buyers runs out.
- Reporting counts instead of ratios. "We got more clicks" means nothing without the impressions and spend behind them.
- Mixing attribution windows. Comparing a seven-day click-attributed result with a one-day view-attributed one, usually without noticing.
Every metric on this page has a glossary entry with a worked example, and the calculators implement the formulas; the funnel guide explains which stage each metric belongs to.
