A testing framework, and creative fatigue

Most people do not test, they fiddle. They change the creative, the audience and the budget in the same week, see a different number, and have no idea which change produced it. A framework is just a rule about what you are allowed to change at once.

Test one thing, in this order

1. 01The offer. The largest effect and the one most often skipped, because it lives outside the ad account.
2. 02The angle. Which argument lands. Different scripts, same production quality.
3. 03The execution. Hook, format, length, within a winning angle.
4. 04Everything else: placements, budget shape, bid settings. Real but small, and worth nothing if the first three are wrong.

Working in that order matters because an effect at the top swamps everything below. Testing thumbnails on an offer nobody wants produces a very precise measurement of nothing.

A test written down before it runs

The discipline is not complicated; it is just that almost nobody writes it before starting. Four lines, decided in advance.

• What is changing: angle A against angle B, same offer, same audience, same production quality.
• What is being measured: cost per enquiry, not click rate, not video views.
• What counts as enough: a stated number of enquiries across the pair, not a number of days.
• What happens next in each outcome: if A wins, it becomes the control and B is replaced by angle C. If they tie, both stay and the next test is the offer rather than the angle.

Writing the last line in advance is the part that matters. Afterwards you will find a reason to keep whichever advert you are fond of, and the written rule is the only thing that stops you.

Enough data to decide

Decide what counts as enough before you launch, and write it down. For most small advertisers the honest threshold is uncomfortable: more conversions than a small budget produces in a week.

• If you cannot reach a decisive number of conversions, judge on a step closer to the top of the funnel, and accept it is a weaker signal rather than pretending it is not.
• Do not judge inside the first few days. Delivery is still learning and early numbers reliably mislead in both directions.
• Do not stop a test because the first day looked bad. That is the most expensive habit on this list.
• Do not restart a test because you edited the advert. An edit resets learning, so the days before it are no longer part of the same test.

Creative fatigue

Every advert here has a lifespan. The same people see it repeatedly, it stops being new, and performance declines. This is not a fault in the advert and it is not something you can fix by raising the bid.

• Watch cost per conversion drifting up while nothing else changed. That is the usual first sign.
• Check how often the same person has seen it. Rising frequency with falling results is fatigue rather than a broken advert.
• Watch the click rate specifically. Fatigue usually shows there before it shows in cost, because people stop stopping before they stop buying.
• Rotate the angle, not just the edit. A new cut of the same argument fatigues almost immediately, because the argument is what people are tired of.
• Keep a queue. The time to make the next advert is while the current one is still working.

Retiring an advert that is still profitable feels wrong, and waiting until it is unprofitable means spending the decline. Rotate on the trend rather than on the cliff.

Fatigue or saturation

Two different things look identical on the chart and need opposite responses, so it is worth separating them.

• Fatigue: the same people have seen this advert too often. Frequency is up, click rate is down, and a genuinely different advert fixes it.
• Saturation: you have reached most of the people worth reaching in this audience. Frequency is up, click rate is flat, and a new advert changes nothing because the audience is the constraint.

The tell is the click rate. If new creative lifts it, you had fatigue. If three genuinely different adverts all land at the same cost, you had saturation, and the answer is a wider audience or a different channel rather than a fourth advert.