Ad budget calculator

Working backwards from the outcome

Budgets set by picking a round number and hoping produce campaigns that cannot succeed at any level of execution. Starting from the outcome makes the arithmetic of the plan visible before any money moves, and it frequently kills the plan, which is the point.

• The chain is short: conversions need clicks, clicks need money. Your conversion rate sets the first link and your click price sets the second.
• Deriving a daily cap matters because that is the field the platform actually asks for. A total budget with no daily figure tends to be spent in the first week.
• The cost per conversion that falls out is the number to argue about. If it exceeds your gross profit per customer, no amount of budget makes the plan work.
• Everything here is media only. Agency fees, creative production and tooling sit on top, and for small budgets they can exceed the media itself.

The exact sums this page runs

Clicks needed is named separately because it is the step where a plan usually falls apart: seeing that a modest conversion target requires four thousand clicks tends to end the conversation faster than any budget figure.

• Clicks needed = (Conversions you want ÷ Landing page conversion rate) × 100. Conversions divided by the rate that produces them, scaled back out of percentage points.
• Total budget = Clicks needed × Expected cost per click. Media cost for the whole run, before agency fees, creative production or tooling.
• Daily budget = (Clicks needed × Expected cost per click) ÷ Days the campaign runs. The cap to type into the platform, which is the number the campaign settings actually want.
• Cost per conversion = (Expected cost per click × 100) ÷ Landing page conversion rate. What each conversion costs at these assumptions. Compare it against gross profit per customer before spending anything.

Getting the three assumptions from somewhere real

The output is only as good as the click price and the conversion rate you feed it, and those are the two numbers people invent. Here is where to find each of them when you have little or no history.

1. 1Take cost per click from your own delivery if you have any at all, even a small test. Thirty real clicks beat any published benchmark.
2. 2With no history, use the platform’s own forecasting tool for your specific keywords, audiences and country, and then add a margin, because forecasts skew optimistic.
3. 3Take the conversion rate from paid traffic only, and ideally from the exact landing page. A site-wide rate includes returning customers and direct traffic and will always be too high.
4. 4With no page live, assume something conservative and treat the whole calculation as a feasibility check rather than a forecast.
5. 5Check the answer against the platform’s learning requirements. Meta needs roughly fifty optimisation events a week per ad set to leave the learning phase, so a budget delivering ten conversions a week will keep restarting learning and underperform.
6. 6Re-run the numbers a fortnight in with real delivery data. The first version of this calculation is a hypothesis, and the second one is a plan.

The budget floor nobody mentions

There is a level below which paid advertising does not work at all, and it has nothing to do with how good the campaign is. Optimisation systems need a minimum flow of conversions to learn from, and below that flow they behave unpredictably.

• A useful floor is a daily budget worth several target conversions, so the system gathers enough signal to optimise rather than guessing.
• Splitting a small budget across many ad sets is the classic mistake. Each one learns separately, so five underfed ad sets perform worse than one properly fed one.
• Long sales cycles make this worse, because the optimisation signal arrives days or weeks after the click. Optimising towards an earlier proxy event is the usual workaround, with the usual risk of optimising for the proxy instead of the sale.
• If the honest budget floor is beyond what you can commit, the correct decision is often to not run paid at all this quarter rather than to run it underfunded and conclude the channel does not work.