Pott Candles · testing review · 28 October 2026

What we've learned about testing, and what we do next

Two questions from our call. When an ad wins in testing, does it win again in the main campaign? And does adding new ads to the main campaign knock it off course? We went through every Pott ad made in 2026, just over 1,000 of them, to answer both.

6 of 9test winners held up after moving into the main campaign

0ads that did badly in testing and then did well in main

+10% vs +9%main campaign cost per sale after new ads went in, vs when none did

In one minute

Testing works as a filter. A good result in testing usually holds in the main campaign. A bad result has never come good.
Testing does not pick the single best ad. How much Meta spends on each ad in a test is not a ranking. Ads that got little spend have not been judged yet, and some of them went on to be our best.
Adding new ads to the main campaign is safe, in any number. It does not measurably upset the campaign, or improve it.
What we do not know yet: whether moving only proven winners beats putting new ads straight in. We have only two clean examples. That is the main thing to test, and January is the time to do it.

What the data says

Eight things we found

A good test result usually carries over

Nine winners have been moved into a main campaign one at a time and run long enough to read. Six held or improved, and so did the ad set around them. Two looked worse, but only in the mid-September week when the whole ad set was struggling. One was never given any spend.

707671725813875852855858

green = held · amber = unclear (bad week) · grey = never fed. Ad 707: £45 per sale in testing, then the biggest ad of the summer; its ad set grew from £3.1k to £8.5k a week at around £41 per sale.

A bad test result has never become a good main ad

Not once in 2026. In April, every ad that was £46 a sale or worse in testing stayed bad, or got no spend, once moved.

How much an ad gets spent on in testing is not a ranking

In April we moved a whole test over. The order Meta spent on the ads in testing and the order it spent on them in main were unrelated. Ads 384 and 385 got just £26 and £15 in the test, then became the main campaign's two top spenders.

So an ad with under about £100 in a test has not been judged. An ad with £200 or more at a bad cost per sale has, and has never recovered.

The same ad can do very differently in a different ad set

Ad 852 ran at £21 per sale in Evergreen Ad Set 1 and £101 in the Seasonal set a fortnight later. Ad 420 (the Flora subscription video) has been £23 in one ad set and £57 in another. Where an ad goes matters as much as which ad it is.

Moving an ad on its original post gives it a head start

New ads on a post that already had 50 or more reactions and comments reached £100 of spend 86% of the time, against 32% for brand-new posts. Part of that is because we only reuse posts from ads that already worked, so treat it as a guide, not a rule. Strong existing ads did not crowd the new ones out.

Adding new ads does not knock the main campaign off course

We looked at the two weeks before and after 24 batches of new ads, from a single ad up to 44, and compared them with two-week spells in the same campaigns when nothing was added. Cost per sale moved the same amount either way, and the day-to-day ups and downs were the same. Batch size made no clear difference.

After new ads went in+10%

When nothing was added+9%

Median change in the main campaign's cost per sale, two weeks after vs two weeks before.

Testing gets ads funded more than it makes them better

Since June, 220 ads went straight into the main campaigns without a test, and half of them never got £100 of spend. Of the 11 that came through testing, only one was ignored. Once ads did get proper spend, results were similar either way: £45 per sale straight in, £41 tested.

The Fir winners have not been read yet

The five Fir winners were copied into the Evergreen Christmas ad set on 27 October, the day after the campaign budget moved to £2,000 a day. The Christmas ad set dropped from £3,547 a day to around £500 to £700, so the Fir copies have had very little spend. That tells us about the budget, not the ads.

What we still do not know

Is moving proven winners better than putting new ads straight in?

Almost everything we have added this year has gone straight in. Proven winners have been moved into a running main campaign on their own only twice (Cook Candles campaign, 4 Aug, £48 → £40 per sale; Roses May 2026 campaign, 14 Aug, £32 → £38 per sale). One went better, one went slightly worse. That is not enough to decide either way, so we will not change the set-up on the strength of it.

How we read results from now on

Five rules

Judge a moved ad on two numbers, not one: its own cost per sale, and the ad set's cost per sale over the same weeks. A top ad drifting to £50 while the ad set grows at £41 is a success.
Under £100 in a test means not judged. Give it one more go in another ad set before dropping it.
£200 or more at a bad cost per sale means judged. Drop it.
Move ads on their original post, so they keep their likes and comments.
Read a moved ad after 7 days, and the campaign after 14. One bad week is a question, not an answer.

Next steps

What we test now, and what waits for January

Now, through ChristmasLow risk. No changes to how testing works during peak season.

Keep the current testing set-up. It is doing its job as a filter, and peak season is the wrong time to change it.
Run the Flora test. Move all four Flora ads from Test 3 into Gifting together, on their original posts, and pause them in Test 3. Not the Christmas ad set, where the budget is too tight to read anything. Read the ads after 7 days and the campaign after 14.
Give the Fir winners a fair read. Either give the Christmas ad set its budget back, or pause the Fir ads in the test campaigns so the spend follows them into main. Read them from 4 November.
Keep adding fresh creative to main as needed. Nothing in the data says it hurts.

January to MarchLower spend and cheaper sales, so a bad week costs far less than in November or December.

Run the comparison we have never run. Alternate between moving only proven winners and putting new ads straight in, in the same campaign, three or four rounds of each, and read every round the same way. This answers the question above.
Give every test ad a fair read. Aim for around £100 on each ad before calling it, by putting fewer ads in each test or setting spend floors.
Repeat the April "move the whole test" once. It is our only example, and one is not enough.