Somebody declares a winner after forty clicks and nine conversions. Two months later the winner is losing and nobody knows when that happened. The test was never capable of answering the question it was asked.
With one ad there is nothing to compare, so every claim about performance is a claim about a single sample with nothing beside it. Three variations in a group costs nothing extra to run: the same budget buys the same clicks, spread across three messages instead of one.
Two ads that differ in headline, description and landing page will produce a winner and no explanation. You will not be able to carry anything from it into the next test, which is the only reason to test at all.
We read your search terms report and send back the list of what to exclude, in the order that saves the most. One email, no call.
The smaller the true difference between two ads, the more traffic it takes to see it. A difference of half a percentage point in click rate needs thousands of impressions per variant, not hundreds. Calling it early is not fast, it is guessing with extra steps.
A brilliant ad pointing at a page that does not repeat its promise loses the visitor in the first two seconds, and the ad gets blamed. Landing page experience is also one of the inputs Google documents for ad quality, so the page moves your costs as well as your conversions.
This is teaching material and our own reading of public documentation, not advice for your specific account. Check anything important with your own specialist before you act on it.
Run the calculator, then tell us what it showed. We reply with what to fix first and why. No call, no pitch.