ToolSahla, Tools Sahla

I tested an ad with a small budget and got only 3 clicks: why should I not trust the rates from such small numbers?

Because with small counts every single event moves the rates a long way. One extra click lifts CTR by a third, and one order takes CPA from 'Undefined' to the entire budget. The calculator computes exactly, but it runs no statistical test, so it cannot tell you whether a difference is real or luck.

Published · 3 min read

Try it yourself: one event more or less

Your ad was shown 1,000 times on a budget of 100 AED, and got 3 clicks and no orders. Enter the figures, then change one event in each direction:

Worked example

As it happened, 3 clicks and no order: CTR 0.3%, CPC 33.3333, CPA 'Undefined'.

One click fewer, so 2 clicks: CTR 0.2%, CPC 50.

One click more and one order, so 4 clicks and an order: CTR 0.4%, CPC 25, CPA 100, conversion rate 25%.

Look at the range: CTR between 0.2% and 0.4%, the best case double the worst, and the whole difference is two clicks. CPC between 25 and 50. And a 25% conversion rate looks wonderful, but it is one order from 4 clicks; had that order not come, it would be zero. None of these numbers can tell you the ad is good or bad.

What a small test can tell you

  • You can state what happened: you spent 100, got 3 clicks and no order. That is a fact.
  • You cannot say this ad's CTR is 0.3%, because the next test could give double or half that by chance alone.
  • You cannot compare an ad with 3 clicks and one with 5 and call the second clearly better.
  • You can rule an ad out only if it is so weak that a few clicks would not rescue it, which is rare with counts this small.

What to do instead of judging early

  1. Enter the figures in the Campaign performance calculator as they are, then recalculate with one event fewer and one more, as above. If your conclusion changes between them, the numbers are too small to judge.
  2. Decide before the test how much you will spend and how many clicks or orders you need before deciding, and do not change the rule halfway.
  3. If the budget cannot stretch to a bigger test, run fewer ads at once so each gets a larger share.
  4. Before any test, find out how many orders you need to cover the spend with the Break-even calculator, so you know what size of result is worth measuring.

Why the calculator gives no statistical verdict

It is built to compute five precisely defined metrics from four figures, exactly and without flattering them. Statistical significance is a different question that needs extra assumptions and data. What the calculator does honestly is show the numbers as they are, with 'Undefined' when there is no order, so you never see a made-up CPA conjured from nothing. Once enough orders build up, use the ROAS calculator to see whether the return covers the spend.

Common questions

How many clicks do I need before I can trust the rate?

There is no single number, and the calculator does not set one. A practical rule is to keep going until changing one event no longer flips your conclusion.

Should I stop the ad because 100 brought no order?

If that is all you can spend, the decision is a business one. On the numbers alone, no order from 3 clicks does not prove the ad failed.