CPA calculator

Ad spend is the media cost for the period, in one currency. Nothing here converts between currencies, and an average taken across two of them without a stated rate is a number nobody can check.

Conversions is the count of one action, defined the same way across the whole period. This is where most of the damage happens. A platform will happily report purchases, leads, add-to-carts and view-through conversions in the same column if the account is configured that way, and an average cost across that mixture describes an outcome that does not exist. Pick the action the business actually pays for, count only that, and run the calculator again for the next one.

Value of one conversion is optional, and it is the field that turns a cost into a judgement. Use the gross value the business keeps — revenue less the cost of goods — rather than the order total. Entering the order total makes the contribution line look comfortable at any margin, because it sets an acquisition cost against money that partly belongs to a supplier.

The first output is the cost itself: spend divided by conversions. The second is what one conversion leaves after that cost, which is the line that decides whether the campaign is worth running. The third restates the same two numbers as a ratio, so the figure can be compared with the return the rest of the account reports.

The second and third outputs are the same fact in two forms, and people reach for different ones. A business that thinks in unit economics reads the contribution: this conversion cost this much and left this much behind. A business that thinks in media efficiency reads the return. Neither is more correct, and having both on the page stops an argument where two people are describing the same campaign in two currencies of thought.

It flattens a distribution. An account with a mean acquisition cost sitting neatly on target can be made of one campaign buying conversions cheaply and another buying them at three times the price. The average is the arithmetic of the two, and it is not a description of either. Segmenting before concluding is the single highest-return habit with this metric.

It ignores what the conversion was worth. Two campaigns with identical acquisition costs, one bringing customers who order once and one bringing customers who order for two years, look the same in the month they ran. Only a value the platform cannot see separates them.

It moves when the definition moves. Adding a second conversion action to an account lowers the reported cost immediately, with no change in the media at all, because the denominator grew. A period-on-period comparison across a definition change is a comparison of two different metrics.

It rewards harvesting. The cheapest conversions are almost always the ones that were going to happen anyway: branded search, retargeting a cart, an audience of existing customers. An account tuned purely on this number drifts toward those and stops reaching anyone new, and the cost stays beautiful right up until volume stalls.

The habits that make it useful are the same ones that make any average useful.

Set it against a ceiling you computed, not a benchmark you read. The maximum you can pay for a conversion is the gross value it brings minus the contribution you need from it. That ceiling comes out of the business, not out of a study about a different one.

Segment before you conclude. Prospecting, retargeting and branded activity have structurally different costs, and blending them gives a figure that describes none of them.

Hold the definition still. Same action, same attribution window, same counting rule, for every period being compared.

Watch it against volume. The interesting question is rarely the cost at today's spend. It is what the cost does as spend rises, because every account has a point where the next conversion costs more than the last, and finding that point is most of the job.

Start with the count rather than the cost. Confirm the conversion fires once per outcome, that duplicates are not being recorded, and that the action being counted is the one the business cares about. A tracking fault inflates the cost exactly the way a media problem does, and it is cheaper to rule out.

Then work on the conversion rate rather than the bid. The cost is spend divided by conversions, so a landing page that converts more of the same traffic lowers it without touching the media at all. This is usually the largest single lever available and the one most often skipped, because it belongs to a different team.

Then narrow the buy. Removing the placements, terms and audiences that consistently convert least lowers the average without lowering the reach that matters.

Then, and only then, move budgets — against the ceiling the value field implies, rather than against the figure somebody else published.