ACoS and ROAS summarize the outcome of advertising. They do not explain it. When ACoS rises, the cause could be more expensive clicks, fewer clicks turning into orders, smaller orders, or a mix of all three. The three funnel metrics in this guide, cost per click, click-through rate and conversion rate, break the outcome into stages so that you can see which one moved.
Three formulas
CPC = spend ÷ clicks
CTR = clicks ÷ impressions × 100%
CVR = attributed orders ÷ clicks × 100%
Cost per click (CPC) is the average price paid for a click over the period. Individual clicks cost different amounts, so this is a period average, calculated from totals.
Click-through rate (CTR) is the share of ad impressions that led to a click. It is a rough signal of how relevant and appealing the ad looked to the people it was shown to, in the placement where it appeared.
Conversion rate (CVR) is the share of clicks that were followed by an attributed order within the attribution window. It reflects the product page, price, reviews, delivery promise and how well the search intent matched the product.
Each formula has a denominator that can be zero. No impressions means CTR is N/A; no clicks means CPC and CVR are N/A. None of these should be shown as 0%.
A fictional funnel
The numbers below describe an illustrative campaign over one month.
| Metric (fictional) | Value | How it is calculated |
|---|---|---|
| Impressions | 40,000 | Counted ad views |
| Clicks | 320 | Counted ad clicks |
| CTR | 0.80% | 320 ÷ 40,000 × 100% |
| Spend | 256.00 | Total cost of clicks |
| CPC | 0.80 | 256.00 ÷ 320 |
| Attributed orders | 32 | Orders credited to ads |
| CVR | 10.0% | 32 ÷ 320 × 100% |
| Attributed sales | 960.00 | Sales credited to ads |
| Average attributed order | 30.00 | 960.00 ÷ 32 |
| ACoS | 26.7% | 256.00 ÷ 960.00 × 100% |
How the stages connect to ACoS
When all figures come from the same report and period, ACoS can be rewritten in terms of the funnel:
ACoS = CPC ÷ (CVR × average attributed order value)
For the funnel above: 0.80 ÷ (0.10 × 30.00) = 0.80 ÷ 3.00 ≈ 26.7%, the same answer as dividing spend by sales directly. The identity shows the three levers that move ACoS: the price of a click, the share of clicks that convert, and the value of each converted order.
| Scenario (fictional) | CPC | CVR | Avg order | ACoS |
|---|---|---|---|---|
| Baseline | 0.80 | 10.0% | 30.00 | 26.7% |
| Clicks cost 25% more | 1.00 | 10.0% | 30.00 | 33.3% |
| Conversion improves | 0.80 | 12.5% | 30.00 | 21.3% |
| Larger orders | 0.80 | 10.0% | 36.00 | 22.2% |
| CTR doubles, nothing else moves | 0.80 | 10.0% | 30.00 | 26.7% |
The last row is worth a second look. CTR does not appear in the ACoS formula. If the ad earns twice as many clicks at the same CPC and CVR, spend and attributed sales both double and ACoS stays where it was. CTR mainly decides how much traffic a given number of impressions produces, which matters for scale, but it does not by itself make each click cheaper or more likely to convert.
Diagnosing a change, stage by stage
When ACoS moves between two periods, compare each stage before deciding what to do. If CPC rose while CVR held steady, the auction became more expensive, and the question is whether higher bids or new competition explain it. If CPC held steady while CVR fell, look at the product page instead: a price change, a lost offer, fewer reviews, slower delivery or a stock problem. If both held steady but order value dropped, check whether shoppers switched to a smaller size or a cheaper variant. Each answer points to a different fix, and only one of them involves bids.
Impressions and clicks are not people
Funnel counts are events, not individuals. That sounds obvious, but it is one of the most common sources of mistaken conclusions.
- An impression is an ad view. One shopper may see the same ad on several searches or pages in a day, and each view counts.
- A click is an interaction. The same shopper may click twice, compare products, leave and come back.
- An attributed order can contain several units, and attribution rules decide which ad gets credit when a shopper interacted with more than one.
- Sessions are visits. On the seller side, sessions count visits to a product page from all sources, not unique buyers.
So “320 clicks” does not mean 320 people visited, and “10% CVR” does not mean one in ten visitors bought. It means 32 credited orders followed 320 clicks.
Seller reports also contain a conversion-style measure, unit session percentage (units ordered ÷ sessions). It covers all traffic, including shoppers who never saw an ad, and counts units rather than orders. It is a useful number, but it is not comparable with ad CVR.
Common mistakes
- Averaging rates. Average CPC, CTR and CVR should be calculated from period totals. Averaging daily percentages gives quiet days the same weight as busy ones.
- Trusting small denominators. A CVR of 0% on 12 clicks, or 25% on 8 clicks, says very little. Wait for enough clicks before drawing conclusions.
- Ignoring attribution delay. Clicks are recorded quickly; orders credited to those clicks can arrive days later. CVR for the latest days is usually understated.
- Comparing unlike placements. CTR varies with ad type, placement and device. Compare like with like.
- Mixing orders and units. CVR here uses attributed orders. Swapping in units changes the metric.
Try it
Enter impressions, clicks, attributed orders, spend and sales into the Ads Metrics Calculator at the free tools to reproduce the funnel above. The synthetic advertising demo shows CTR, CPC and ACoS for sample campaigns, with “Explain metric” on each value. Next, read What Is ACoS? or Seller Sales vs Ad-Attributed Sales, and check terms in the glossary.