A responsive search ad is not published and forgotten: it is cultivated. Every piece accumulates data, Google favours the ones that work, and the weak ones sit there taking up a slot and dragging the combinations down. This lesson closes the module with the method for knowing which pieces perform and how to replace them without breaking what already works.
Where you see what works
Inside an ad → View asset details. Two views:
Assets
Every headline and description with its impressions and a performance rating: Low, Good, Best (or Learning / Pending if there is no data yet). It is relative: it compares each piece with the others in the same ad.
Combinations
The combinations of headlines and descriptions that have been shown most. Useful for checking that the pairings make sense and that on mobile (two headlines) the message is still complete.
One important limitation: the ratings need volume (thousands of impressions) and they are calculated on impressions, not on conversions. A «Best» piece is the one Google shows most; that is almost always the one that performs best, but not always.
When to change and when to wait
| Situation | What to do |
|---|---|
| Piece on «Learning» or «Pending» | Wait. With no data there is no decision |
| «Low» piece with volume (thousands of impressions) | Replace it with a new one on a different angle |
| «Good» piece | Leave it; it is the backbone of the ad |
| «Best» piece | Don't touch it. Use it as a model for writing new ones |
| Everything «Low» or CTR falling across the board | The problem is not the pieces: review the ad group (intent), the page or the competition |
The healthy cadence: review every 3-4 weeks per ad group with volume. Every few days is noise; every six months is neglect.
The golden rule: one piece at a time
When you replace a piece, the rest of the ad keeps its learning. If you rewrite ten headlines at once, Google has to relearn almost everything and performance drops for weeks. That is why optimisation is done continuously and in small doses: one or two weak pieces per round, replaced by candidates on an angle that is not yet covered (go back over the eight angles in lesson 2).
Three common mistakes:
- Deleting the ad and creating a new one «to start from scratch». You throw away the history.
- Copying the best piece with a synonym as a new candidate. It is the same piece; nothing is learned.
- Changing pieces without writing it down. Three weeks later you won't know what changed. Keep a record: date, old piece, new piece, reason.
Testing between ads: carefully
Having two or three RSAs per ad group is already a continuous test: with rotation set to «Optimize», Google gives more impressions to the one that performs better. To genuinely compare two approaches (price versus guarantee, say), Google Ads has Experiments (ad variations at account level): it applies a change to a percentage of the traffic and compares rigorously. That is the clean way to do it; the dirty way (switching ads on and off and looking at CTR) mixes different periods and misleads you.
What to look at besides CTR
CTR tells you whether the ad wins the click. But the goal is the conversion. A piece that raises CTR and lowers the conversion rate (it over-promises) is worse than a modest one that brings in the right buyer. When you evaluate a change, look at the whole chain: CTR → conversion rate → CPA.
An ad is never finished
Competitors change their ads, searches evolve, your offers expire. The ad that was best in January is mediocre by July. The only way to keep the advantage is the cycle: read the report → replace the weak → wait → read again. Done with discipline, it is the cheapest compound improvement in Google Ads: the same budget buys a little more every month.
💡 Ninja trick: this cycle is the essence of «The eternal ad», the ebook that accompanies our RSA Optimizer. The script reads the asset report of every ad group, identifies the weak pieces with volume, generates candidates with AI covering the empty angles (drawing on the real search terms and on the relevance measured by the QS Analyzer) and leaves them in a spreadsheet: you approve with a checkbox and the script swaps them in one at a time, with a record of every change. The ad cultivates itself.
What you should remember
- The asset report tells you which pieces perform (Low / Good / Best); it needs volume and it measures impressions, not conversions.
- Replace the Low ones with volume with new angles; don't touch the Best ones; wait when there is no data.
- One or two pieces per round, every 3-4 weeks, with a record.
- Judge by the whole chain, not by CTR alone.
- An ad is a crop, not a delivery.
End of Module 4. In Module 5 we move on to the money: budgets and bids — what target CPC, CPA and ROAS mean, how the daily budget works and where NOT to touch.