How to optimise Google Ads ads when you have too few conversions to judge

This is our argument, not Google's. A test judged on conversions alone needs enough conversions per ad to separate them, and on an ad group with a handful a month that can take longer than the offer, the season or the landing page will last. While you wait, every ad in the test keeps taking a share of the budget, the weak ones included.

Google's own testing advice frames the threshold in impressions. It says to establish a testing threshold and stick to it, and to wait until you have served enough impressions to be confident in your results. It also says looking at clickthrough rate or conversion rate alone can be misleading. So the question for a small account is not whether to wait, but what to read while you do.

On the 'Optimise' ad rotation setting, Google says your ads are optimised for clicks in each individual auction, and that as data is accumulated, ad serving will become weighted more heavily in favour of the ads that look statistically likely to perform better. Ads expected to attract more clicks, and conversions if you're using a Smart Bidding strategy, are delivered more often into the ad auction than other ads in the ad group. If you're using Smart Bidding, Google says Google Ads will automatically use the 'Optimise' setting.

Google's testing guidance draws the conclusion itself: if the system is showing an ad more often than another, it can mean that it's better at winning impressions in the auction itself, which is often a strong reason to prefer that creative over others.

Our reading: that makes each ad's share of its ad group's impressions the cheapest signal a small account has. It is built from every auction the ad group entered, not only the ones that ended in a conversion, and Google has already done the weighting.

Google says it's possible for ad serving to remain relatively even on 'Optimise', for example if ads in an ad group have similar performance, or if the ad group doesn't receive a significant amount of impressions and clicks for some period of time.

Our reading: an even split is not a verdict that the ads are equal. On a quiet ad group it may only mean the system has not seen enough yet. Check the ad group's volume before reading an even split as a tie.

This ladder is our practice, not Google's. Take the highest rung the ad group's volume can support, and write down which one it was.

First, conversions per ad, when there are enough of them to tell the ads apart. Second, each ad's share of the ad group's impressions under 'Optimise', the signal above. Third, clickthrough rate, read beside the serving share and never instead of it, because Google warns that responsive search ads help you qualify for more auctions, and there are many instances where you might end up serving impressions in a low clickthrough rate placement that you might not have qualified for before. A decision taken on the second or third rung is a narrowing of the test, not a winner, and we label it that way.

Ad Strength is not on the ladder. Google says it measures the relevance, quality and diversity of your ad assets, and that a higher ad strength will help you to maximise your ad's performance. Our reading: it describes how the ad is built, not how it has done, so it cannot tell you which of your ads is winning.

This part comes from what we saw on one account we manage, so it carries no figures. When ads were rebuilt for a new landing page, variants the system had already been showing rarely were taking impressions again in the weeks after, and the variants it had favoured were sharing the budget with them.

Google does warn that removing or editing any headlines or descriptions can change the ads that are serving and may impact your performance. Our practice, from that experience: before rebuilding an ad group's ads, pause the variants Google has been starving, on its own serving share and click rate, so the rebuild starts from the ads the system was already backing rather than from the whole field.

Google says ad variations allow you to easily create and test variations of your ads across multiple campaigns or your entire account, and that they are best used when you want to test one change across multiple campaigns or your entire account, while custom experiments are best to test multiple simultaneous changes on a smaller scale.

Our reading: for a small account that is the useful property. One change, such as a call to action, tested across every campaign that uses it gathers impressions from all of them at once. Google's own advice points the same way: limit how many elements you test within your ads, and make different variations similar enough that you learn from their differences.

Our practice, set out as a checklist. Check the ad group's volume before reading anything. When conversions are too few to separate the ads, read each ad's share of impressions under 'Optimise' rather than its conversion rate. Use clickthrough rate only beside that share. Pause the starved variants before a rebuild, not after it. Test single changes across campaigns rather than many changes in one ad group. And with every decision, record which rung of the ladder it rested on, so the next person knows how sure we were.