Advertising metrics become useful when every number has a definition, a denominator, and a decision attached to it. A dashboard can display impressions, clicks, conversions, revenue, and cost while still leaving the team unsure whether the campaign helped the business. The problem is often not a lack of data. It is a lack of agreement about what the data represents.
This guide explains common display-advertising calculations using one hypothetical campaign, then shows how to connect them to a review. All example values are invented for arithmetic and should not be treated as market benchmarks. The purpose is to help you ask whether a comparison is fair, whether a result is complete, and whether a metric is being used for the job it can actually perform.
Begin with a consistent example
Define the inputs before calculating
Imagine a campaign with $2,000 in media spend, 500,000 reported impressions, 1,000 reported clicks, 40 recorded purchases, and $6,000 in attributed revenue. Before calculating anything, define the reporting period, currency, time zone, attribution settings, and whether canceled orders or refunds have already been removed.
Keep the source of each number visible. Media spend may come from the buying platform, visits from analytics, and completed orders from a commerce system. These records do not necessarily count identical events. A click can occur without a successfully measured visit, and a purchase can be observed under different attribution rules.
Write down which source governs which decision. The payment record may govern settled revenue, while the platform record may govern media billing. A consistent review is more defensible than selecting whichever dashboard happens to show the most favorable number.
CPM describes the cost of reported impressions
CPM means cost per thousand impressions. Calculate it as spend divided by impressions, multiplied by 1,000. In the example, $2,000 divided by 500,000, multiplied by 1,000, produces a $4 CPM. That describes the relationship between the stated spend and the stated impression count.
It does not tell you whether the audience was relevant, whether the ad was noticed, or whether the destination was useful. A lower CPM can be attractive when the opportunities are comparable, but different placements and audiences can make a simple comparison misleading.
Ask whether all relevant costs are included. Platform fees, creative costs, verification charges, or other expenses may sit outside a headline media number. Decide whether the review needs media-only CPM or an all-in cost view, label the calculation clearly, and avoid switching definitions between campaigns.
CTR describes clicks relative to impressions
Click-through rate is clicks divided by impressions, expressed as a percentage. The example has 1,000 clicks from 500,000 impressions, giving a CTR of 0.2%. Google’s CTR definition uses that relationship between clicks and impressions.
CTR can help investigate whether a creative and placement are generating click activity. It cannot tell you by itself whether those clicks are relevant or profitable. A misleading action label could attract clicks that do not represent useful interest. A campaign designed for another objective may also need evidence beyond direct click behavior.
Compare like with like where possible. Device, placement, audience, offer, and timing can influence what the ratio means. Treat a difference as a prompt for investigation rather than automatically crediting the headline. Inspect the destination and later outcomes before deciding that the highest CTR creative should receive more budget.
CPC describes the cost of reported clicks
Average cost per click is spend divided by clicks. The example’s $2,000 spend and 1,000 clicks produce a $2 CPC. This is a useful cost relationship, but it does not establish the value of a click. That value depends on what happens afterward and on the economics of the offer.
A campaign with a higher CPC can still produce a lower cost per qualified outcome when its traffic is more relevant. Conversely, cheap clicks can be expensive when almost none lead to useful activity. Keep click cost connected to the campaign’s primary objective.
Review discrepancies between clicks and measured sessions without assuming one system must be wrong. Redirects, blocked measurement, interrupted page loads, and counting differences can create gaps. Establish whether the gap affects the decision you are making, then investigate the implementation with the relevant technical owner.
CPA requires a clearly named action
Cost per acquisition or cost per action is spend divided by the counted outcomes. In the example, 40 purchases from $2,000 in spend yield a $50 cost per purchase. The label matters: replacing “purchase” with the vague word “conversion” can hide important differences.
For a lead campaign, calculate separate costs for inquiry, qualification, accepted opportunity, and sale when the data supports those stages. The lead-generation guide explains why a form submission should not automatically be treated as a qualified business opportunity.
Be careful with incomplete outcomes. Recent inquiries may not have had time to qualify, and recent orders may still be subject to cancellation. Label immature reporting periods and compare cohorts with similar opportunity to progress. A precise formula applied to an unfinished denominator can still lead to a poor decision.
ROAS is a revenue ratio, not profit
Return on ad spend is attributed revenue divided by advertising spend. With $6,000 in attributed revenue and $2,000 in media spend, the example has a ROAS of 3, often written as 3× or 300%. That means three dollars of attributed revenue per dollar of the stated ad spend.
It does not mean three dollars of profit. Product costs, fulfillment, returns, discounts, payment fees, creative production, staff time, and other expenses may still need to be considered. Revenue attribution also depends on the measurement rules used by the reporting system.
Make a separate contribution calculation using the business’s actual definitions. A revenue ratio can help monitor campaigns, but it should not be presented as a full profitability analysis. Keep reported attribution distinct from a causal claim that the campaign created all of the revenue assigned to it.
Conversion rate depends on the denominator
In the example, 40 purchases divided by 1,000 clicks gives a click-to-purchase rate of 4%. If you use measured sessions instead of clicks, the denominator may differ and so will the result. Neither label should be silently substituted for the other.
Specify whether you are counting events, users, sessions, orders, or accounts. A person who purchases twice can affect an event-based calculation differently from a user-based one. A reporting tool may also apply its own counting and attribution conventions. Read those definitions before combining numbers in a spreadsheet.
Use the rate to locate a question, not to supply a diagnosis automatically. A change can reflect audience mix, offer availability, page behavior, measurement changes, or delayed outcomes. Review the full context before declaring that a design change caused the movement.
Turn the calculations into a decision record
Create a review with the primary outcome, supporting metrics, total costs, relevant context, and known limitations. State what changed during the reporting period. Include campaign and creative versions, destination updates, offer changes, and any operational interruptions that could affect the comparison.
Then write the decision in plain language: continue the current plan, investigate a delivery issue, improve the destination, stop an irrelevant placement, or run a focused test. Explain which evidence supports the choice and what evidence would change it. The creative testing guide helps turn uncertainty into a cleaner comparison.
Good measurement does not require every number to improve simultaneously. It requires the team to understand trade-offs without changing definitions midstream. Keep the arithmetic transparent, distinguish early activity from completed value, and use each metric to answer the question it is capable of answering.
