Short answer
An ad budget is worked out backwards from the target: first the contribution an order leaves gives the most you can spend on one order, then budget = target orders × target cost per order gives the figure. If the budget cannot fund the orders the platform needs to learn, use fewer channels.
Everyone in the meeting asks the same question: how much should we spend each month? One person suggests a percentage of last year’s revenue, another guesses what a competitor spends, a third says “let’s start small”. Each leads to a number, but none says whether that number leaves a profit on your orders. The useful answer to how to set an ad budget is to work backwards from the target, using only your own figures. This article builds that calculation in order, with example units; the numbers in the example are round values chosen only to show the formula, not a benchmark for anyone.
The calculation has six steps, and the output of each is the input of the next.
1. Find what an order leaves
The budget starts from the money one order can give to advertising. From the selling price, take off the product cost, shipping and packaging, the payment commission, the marketplace share if there is one and the expected cost of returns; what remains is the contribution. Divide it by the selling price to get the contribution margin. Everything is taken on the same basis, excluding VAT.
Two limits follow. Break-even ROAS is 1 / contribution margin, and the most you can spend on advertising for one order is the contribution itself. We explain why the formula works, and why platform figures and till figures drift apart, in measure before you raise the ad budget; the break-even ROAS calculator applies the same formula to your own product.
2. Choose a target cost below the limit
The limit itself is not a target; at that point advertising only pays for itself. Set the target cost per order clearly below it, because fixed costs such as rent, salaries and the agency fee sit outside the limit. How far below depends on your fixed costs and on how often customers buy again; work out the second from your own sales data, not from a guess.
The table below carries this and the following steps through one example. The amounts are in “units”; they are not a real price or a market average.
| Item | Calculation | Result (units) |
|---|---|---|
| Selling price (excluding VAT) | 800 | |
| Product cost | 400 | |
| Commission, shipping and returns allowance | 200 | |
| Contribution | 800 - 400 - 200 | 200 |
| Contribution margin | 200 ÷ 800 | 0.25 |
| Break-even ROAS | 1 ÷ 0.25 | 4 |
| Most to spend on ads per order | the contribution itself | 200 |
| Target cost per order | chosen below the limit | 150 |
| Target orders per month | 100 | |
| Monthly ad budget | 100 × 150 | 15,000 |
| Average daily budget | 15,000 ÷ 30.4 | about 493 |
3. Write down how many orders or enquiries you need
The budget formula is simple: budget = target orders × target cost per order. If your target is revenue, budget = target revenue ÷ ROAS gives the same result. To try it with your own target and cost, use the ad budget calculator.
Where the sale closes through a form or on the phone, the unit is an enquiry. The value of an enquiry is the contribution per sale times the share of enquiries that become sales. If the contribution per sale is 3,000 units and, by your own records, one enquiry in five becomes a sale, an enquiry is worth 600 units. Choose a target cost per enquiry of 400 units and 10 sales need 50 enquiries, so the budget is 50 × 400 = 20,000 units. Take the close rate from the real records in your CRM.
If you have no history, the cost per order is not known yet. Then the test budget in step 4 comes first, and the calculation is repeated with the cost the test reveals.
4. Set aside a test budget big enough to learn from
The problem with a small budget is not that it spends little but that there is nothing to read in the result. A test budget has to bring in enough orders to decide on.
- Meta: According to Meta, an ad set is labelled “learning limited” when it is unlikely to reach about 50 optimisation events in the week after its last significant edit. At the example cost of 150 units, that is about 7,500 units a week for a single ad set. A monthly budget of 15,000 units is about 3,450 a week, so even one ad set optimising for purchases falls below the line. The answer is to keep the budget in one ad set or to optimise for an event that happens more often.
- Google Ads: A budget is an average daily amount per campaign. Google says spend can reach twice that amount on a given day and 30.4 times the daily budget in a month, so divide a monthly amount by 30.4 to get the daily figure. In Smart Bidding a budget change does not start the “Learning” status, but performance can take about as long as your conversion cycle to settle on the new input.
- Duration: Do not run the test for less time than your buyers take to decide. You will not see the customer who puts something in the basket and buys three days later in a two-day test.
For setting up each platform, see your first Google search campaign and your first Instagram and Facebook campaign; here we stay with the size of the budget.
5. Split it across channels
The main rule is simple: give a channel only as much budget as lets it meet its own test minimum. Two small channels below the line teach less than one large channel above it.
| Situation | Decision |
|---|---|
| The budget barely meets the test minimum of one channel | Keep it in one channel; leave the second for later |
| The budget meets the minimum of both channels | Split by their minimums and track the cost per order for each separately |
| One channel’s next-order cost is nearing the limit | Reduce its share and move the difference to the channel that stays under the limit |
Do not set the split once and leave it; update it from the net sales in your reconciliation table and from the next-order cost per channel. Keep orders from brand searches apart, otherwise the average looks better than it is.
6. When to raise and when to cut
Tie the decision to raise or cut to the cost of the added order; the average ROAS hides it. Write the thresholds beforehand:
| What you see | Decision |
|---|---|
| Added orders cost clearly less than the limit, and the ratio of platform to till is stable | Raise by one step and wait two weeks |
| The cost of an added order has been near the limit for two weeks running | Go back to the previous budget |
| The cost of an added order has passed the limit | Cut, and check measurement first |
| Platform ROAS is rising but net sales have not moved | Do not raise; go back to measurement |
Raise in small steps, changing one thing at a time; Meta may treat a bid or budget change as significant depending on its size. The four measurement questions and the order for raising are in measure before you raise the ad budget; the layout in the Monday report helps with reading the difference each week.
The agency fee is not part of the budget
The ad budget goes to the platform; the agency’s fee is a separate item. Mixing them up causes errors in both directions. If you take the fee out of the budget, less reaches the platform than you think. If you leave the fee out of the calculation altogether, profit looks higher than it is. The right way is to work the budget out from the contribution of an order, as here, and to track the fee as its own line among fixed costs. In our advertising service the budget and the fee are separate items.
Common mistakes
- Setting the budget as a percentage of revenue without looking at the contribution margin.
- Looking at the average cost per order and missing the cost of the added order.
- Slicing a small budget across several channels and getting a readable result from none.
- Comparing a figure with VAT to one without.
- Judging a test before a week has passed, while the platform is still learning.
- Putting the fee inside or outside the budget in the wrong place.
Common questions
What percentage of revenue should go to advertising?
There is no percentage that suits everyone. Your contribution margin sets the ceiling: when advertising reaches your margin as a share of revenue, no profit is left. How far below it you stay depends on your fixed costs and growth target, so work the budget out from orders and cost per order instead of a percentage.
How big should a test budget be?
Big enough to produce a result you can read: multiply your target cost per order by the number of orders the platform needs to learn. On Meta that is about 50 optimisation events in a week. If your budget cannot fund it, use fewer channels or optimise for an event that happens more often.
When should I raise the budget?
When the added orders cost clearly less than the break-even limit and the ratio between platform and till is stable. Take one small step, wait two weeks and look at the cost of the next order again. If platform ROAS rises while net sales stand still, do not raise it.
One thing you can do today: write down the selling price, product cost, commission and shipping of your best-selling product, and fill in the first seven rows of the table with your own numbers. Put the “most to spend on ads per order” you get next to the cost per order in your ad account for the latest period, and those two numbers will often tell you whether to raise or cut.
Sources
This article is general information, not legal, financial or technical advice. Platforms, their rules and the law change over time, so some details may have changed since 11 November 2026 or contain small errors. Before you make a decision for your business, check the current official sources or write to us.