Start with what you can afford to lose
Before any calculation, decide the most you are willing to spend on a first test, even if it produces nothing. Advertising results are uncertain, and a first campaign is partly a way of learning. A budget you can live with is one whose total loss would not hurt the business or cause you to panic and switch things off too early.
This article is general education, not financial advice. All numbers below are simple illustrations, not benchmarks.
The daily budget is a limit, not a promise
Most platforms ask for a daily budget: the average amount you are happy to spend per day. Platforms may spend somewhat more than the figure on some days and less on others, as long as the total over a period stays within the limit they describe. Check the current documentation for how your platform handles this.
Setting a lower budget does not make ads cheaper per click. It only limits how many clicks you can buy. A budget that is too small may run out early in the day and miss the people you wanted to reach.
Work backwards from the result you need
A useful method has four steps.
Step 1: Decide your acceptable cost per result
Take the value of one result and decide how much of it you can spend to get it. If a customer is worth a profit of 80 on a first order and you are happy to spend up to half of that on advertising, your acceptable cost per acquisition is 40. The meaning of CPA is explained more fully in the guide on core paid metrics.
Step 2: Estimate cost per click and conversion rate
Use the platform’s keyword research estimates for click prices, but treat them as rough. For conversion rate, if you have no data, make a deliberately cautious guess. Suppose the click price is about 1.50 and 4 percent of clicks convert.
Step 3: Calculate the likely cost per result
CPA = click price / conversion rate = 1.50 / 0.04 = 37.50.
This is under the acceptable 40, so the plan looks workable, though only just. If the conversion rate were 3 percent, the CPA would be 1.50 / 0.03 = 50, over the limit. Showing how sensitive the answer is to the guess is a good reason to run a test before committing.
Step 4: Decide how many results you need to learn something
A test with one or two conversions tells you very little. Aim for enough results to see a pattern; for illustration, say 10. At a CPA of 37.50 that would cost 10 x 37.50 = 375. You would need about 10 / 0.04 = 250 clicks, and 250 x 1.50 = 375, which matches.
If the test runs for 30 days, the daily budget is 375 / 30 = 12.50.
What if that is more than you can spend?
Suppose you can only afford 200 in total. At 37.50 a result, 200 would be expected to produce about 5 conversions (200 / 37.50 is a little over 5). Five is a thin basis for decisions, since a few extra or missing sales would change the picture a lot. Your options include:
- narrowing the campaign to fewer, better keywords, so each click is more likely to count;
- aiming for an earlier, cheaper action such as an enquiry or sign-up, if that is a real step toward a sale;
- running the test over a longer period at a lower daily rate, accepting it will take more time;
- delaying until the budget is larger.
Spreading a very small budget across many keywords, locations and ads usually gives each too little data to learn from.
Allow for the learning period
Automated bidding systems often take some time and data to settle, and early performance can be worse than later performance. Resist the urge to change settings every day. Equally, set a checkpoint in advance, for example after two weeks, to review results against your acceptable CPA.
Account for everything else
The ad cost is not the whole cost. Remember:
- the time or fees for building the campaign and landing page;
- tools and any creative costs;
- the margin on what you sell, shipping and returns.
A marketing budget for the year should cover these too; an ad budget is one line inside it.
Control the downside
- Set a total cap in your own records as well as a daily budget in the platform.
- Check spend in the first few days, in case something is wrong with targeting or tracking.
- Use negative keywords and tight match types to keep the budget on relevant searches.
- Pause if spend runs with no recorded conversions and you have confirmed that tracking works.
Raising the budget later
Increase spending only after you have a reasonable stretch of data showing results at or below your acceptable cost. Increase in steps and watch whether the cost per result holds, since extra spend often reaches less valuable searches and costs per result can rise. Outcomes vary, and past results do not guarantee future ones.