Pott Candles · testing review, part 2 · updated 28 Oct 2026
What happens to the main campaign in the two weeks after new ads go in?
Campaign-level results only, day by day: 14 days before each batch of new ads was added, 14 days after. Not how the new ads themselves did. Every batch of new ads of any size, from a single ad to 44, added to a main campaign that had been running at least 10 days and spending at least £50 a day, January to mid October. Compared against two-week stretches in the same campaigns when no ads were added.
24batches of new ads (16 untested straight in, 6 mixed, 2 tested winners only)
+10% vs +9%cost per sale two weeks later: with new ads vs with none added
2clean moves of tested winners on their own: too few to judge
Short version
Adding new ads does not measurably unsettle the main campaign. After new ads went in, 31% → 46% of days were "rough" (more than 25% off the earlier cost per sale). In two-week stretches with no new ads at all, the same measure went 39% → 50%. The day-to-day swings are normal for the campaign, not caused by the new ads.
It does not measurably help or hurt cost per sale either. Two weeks after new ads, cost per sale was a median +10% against the two weeks before, with 12 of 24 batches ending more than 10% dearer. With no new ads it was +9%, 7 of 15 dearer. Most batches also arrived with a budget increase and a busier season, so neither number is a clean read on the creative itself.
Batch size makes no clear difference. One or two ads did not land any more smoothly than 20 or more. Bigger batches cost more sales in total, mostly because they went into bigger campaigns spending more.
Moving proven test winners on their own has not been tried often enough to judge. Only 2 clean cases (4 Aug Cook Candles, 1 ad, £48 → £40 per sale; 14 Aug Roses May 2026, 1 ad, £32 → £38 per sale). The Flora move into Gifting is the next one, and it should be read the same way as these.
1 · With and without new ads
The same two-week read, side by side
"No new ads added" means two-week stretches in the same campaigns with no ad created within 14 days either side, stepped a week at a time. They are measured exactly the same way as the batches.
Two-week window
Count
Cost per sale change (median)
>10% dearer
Rough days
No new ads added
15
+9%
7 of 15
39% → 50%
New ads added (all batches)
24
+10%
12 of 24
31% → 46%
untested, straight in
16
+7%
7 of 16
35% → 48%
mixed tested and untested
6
+12%
4 of 6
22% → 33%
tested winners only
2
+0%
1 of 2
17% → 71%
Rough days rise in every row, including the one where nothing changed. That is partly how the measure works: the days before are compared with their own average, so they naturally sit close to it, while the days after are compared with a different fortnight's average. What matters is the comparison between rows, not the rise within a row.
2 · By batch size
Does a smaller batch land more gently?
Ads in the batch
Count
Cost per sale change (median)
>10% dearer
Rough days
1 to 2 ads
4
+13%
2 of 4
27% → 64%
3 to 5 ads
6
+0%
3 of 6
42% → 39%
6 to 12 ads
4
+10%
2 of 4
25% → 34%
13 or more ads
10
+5%
5 of 10
28% → 49%
No clear pattern. Each group holds only 4 to 10 batches, so differences of 10 points either way are within the noise.
3 · All batches on one chart
How far off the before-average, and for how long
Each thin line is one batch. The value is the 3-day rolling cost per sale divided by the campaign's average over the 14 days before, so 1.0x means "same as before" and 1.5x means 50% dearer. The thick orange line is the median of all batches with new ads. The dashed dark line is the median of the stretches with no new ads.
untested ads straight in (16)mixed batches (6)tested winners only (2)median, no new ads (15)
Both medians wander about 10% either side of the before-average. The new-ads line sits a little higher in the second week (around 1.1x against about 1.0x), which is within the noise for this many windows. Lines are capped at 2.2x for readability; hover a line for the batch.
4 · Every batch, day by day
The two weeks before and after each one
Dots are daily cost per sale for the whole campaign. Grey dots are before, coloured dots after. The dashed line is the 14-day-before average and the grey band is within 25% of it. A × at the top is a day with spend and no sales. Tested-only moves first, then mixed, then straight-in, each in date order.
4 Aug · Cook Candles1 ad
£104/d → £79/d · per sale £48 → £32 (wk1) → £40 (2wk)
rough days: 3 of 12 before, 12 of 14 after · not settled by day 13
14 Aug · Roses May 20261 ad
£155/d → £142/d · per sale £32 → £33 (wk1) → £38 (2wk)
rough days: 1 of 12 before, 8 of 14 after · settled day 12
2 May · Evergreen 20263 ads, 1 tested
£175/d → £150/d · per sale £38 → £38 (wk1) → £43 (2wk)
rough days: 5 of 12 before, 7 of 14 after · not settled by day 13
14 Aug · Citronella May 20264 ads, 1 tested
£177/d → £145/d · per sale £41 → £33 (wk1) → £33 (2wk)
rough days: 6 of 12 before, 4 of 14 after · settled day 11
14 Aug · Evergreen (old Tomato vine)12 ads, 3 tested
£426/d → £649/d · per sale £33 → £34 (wk1) → £37 (2wk)
rough days: 0 of 12 before, 4 of 14 after · not settled by day 13
£1,172/d → £1,872/d · per sale £51 → £49 (wk1) → £48 (2wk)
rough days: 3 of 12 before, 3 of 14 after · settled day 13
4 Feb · Evergreen 202613 ads
£499/d → £658/d · per sale £27 → £28 (wk1) → £36 (2wk)
rough days: 1 of 12 before, 8 of 14 after · not settled by day 13
20 Feb · Evergreen 20269 ads
£700/d → £611/d · per sale £41 → £46 (wk1) → £30 (2wk)
rough days: 5 of 12 before, 6 of 14 after · settled day 12
11 Mar · Evergreen 202618 ads
£700/d → £921/d · per sale £29 → £66 (wk1) → £61 (2wk)
rough days: 8 of 12 before, 14 of 14 after · not settled by day 13
17 Apr · Cook Candles5 ads
£74/d → £96/d · per sale £47 → £58 (wk1) → £41 (2wk)
rough days: 4 of 12 before, 8 of 14 after · not settled by day 13
19 May · Cook Candles23 ads
£122/d → £149/d · per sale £52 → £43 (wk1) → £35 (2wk)
rough days: 9 of 12 before, 11 of 14 after · not settled by day 13
17 Jun · Citronella May 202624 ads
£166/d → £148/d · per sale £50 → £35 (wk1) → £40 (2wk)
rough days: 1 of 12 before, 4 of 14 after · not settled by day 13
17 Jun · Evergreen (old Tomato vine)5 ads
£127/d → £205/d · per sale £18 → £25 (wk1) → £24 (2wk)
rough days: 5 of 12 before, 9 of 14 after · settled day 13
14 Jul · Citronella May 202616 ads
£150/d → £150/d · per sale £58 → £52 (wk1) → £57 (2wk)
rough days: 5 of 12 before, 7 of 14 after · settled day 13
14 Jul · Evergreen (old Tomato vine)30 ads
£231/d → £290/d · per sale £23 → £45 (wk1) → £35 (2wk)
rough days: 0 of 12 before, 9 of 14 after · not settled by day 13
5 Aug · Evergreen (old Tomato vine)2 ads
£323/d → £475/d · per sale £27 → £34 (wk1) → £36 (2wk)
rough days: 4 of 12 before, 9 of 14 after · not settled by day 13
15 Aug · Cook Candles1 ad
£85/d → £80/d · per sale £37 → £56 (wk1) → £40 (2wk)
rough days: 5 of 12 before, 7 of 14 after · not settled by day 13
31 Aug · Evergreen (old Tomato vine)9 ads
£720/d → £1,047/d · per sale £35 → £38 (wk1) → £41 (2wk)
rough days: 3 of 12 before, 7 of 14 after · settled day 12
4 Oct · Cook Candles3 ads
£198/d → £208/d · per sale £43 → £35 (wk1) → £38 (2wk)
rough days: 7 of 12 before, 2 of 14 after · settled day 5
4 Oct · Evergreen (old Tomato vine)44 ads
£1,609/d → £2,464/d · per sale £55 → £44 (wk1) → £47 (2wk)
rough days: 4 of 12 before, 2 of 14 after · settled day 5
12 Oct · Cook Candles3 ads
£199/d → £231/d · per sale £42 → £38 (wk1) → £46 (2wk)
rough days: 3 of 12 before, 3 of 14 after · settled day 12
12 Oct · Evergreen (old Tomato vine)8 ads
£2,038/d → £2,958/d · per sale £49 → £50 (wk1) → £52 (2wk)
rough days: 4 of 12 before, 2 of 14 after · settled day 11
The y axis of each panel is capped at three times the before-average so one bad day does not flatten the rest; hover a dot for the exact figures. "Rough days" counts days where the 3-day rolling cost per sale was more than 25% away from the before-average. "Settled" is the first day after which it stayed inside 25% to the end of the window; it gets easier to pass the closer a day is to the end, so late "settled" days mean little.
5 · The numbers behind the panels
Summary table
Date
Campaign
Type
Ads added
Spend/day before → after
Per sale before → wk1 → 2wk
Rough days
Settled day
4 Feb
Evergreen 2026
straight
13
£499 → £658
£27 → £28 → £36
1/12 → 8/14
–
20 Feb
Evergreen 2026
straight
9
£700 → £611
£41 → £46 → £30
5/12 → 6/14
12
11 Mar
Evergreen 2026
straight
18
£700 → £921
£29 → £66 → £61
8/12 → 14/14
–
17 Apr
Cook Candles
straight
5
£74 → £96
£47 → £58 → £41
4/12 → 8/14
–
2 May
Evergreen 2026
mixed
3 (1 tested)
£175 → £150
£38 → £38 → £43
5/12 → 7/14
–
19 May
Cook Candles
straight
23
£122 → £149
£52 → £43 → £35
9/12 → 11/14
–
17 Jun
Citronella May 2026
straight
24
£166 → £148
£50 → £35 → £40
1/12 → 4/14
–
17 Jun
Evergreen (old Tomato vine)
straight
5
£127 → £205
£18 → £25 → £24
5/12 → 9/14
13
14 Jul
Citronella May 2026
straight
16
£150 → £150
£58 → £52 → £57
5/12 → 7/14
13
14 Jul
Evergreen (old Tomato vine)
straight
30
£231 → £290
£23 → £45 → £35
0/12 → 9/14
–
4 Aug
Cook Candles
tested
1
£104 → £79
£48 → £32 → £40
3/12 → 12/14
–
5 Aug
Evergreen (old Tomato vine)
straight
2
£323 → £475
£27 → £34 → £36
4/12 → 9/14
–
14 Aug
Roses May 2026
tested
1
£155 → £142
£32 → £33 → £38
1/12 → 8/14
12
14 Aug
Citronella May 2026
mixed
4 (1 tested)
£177 → £145
£41 → £33 → £33
6/12 → 4/14
11
14 Aug
Evergreen (old Tomato vine)
mixed
12 (3 tested)
£426 → £649
£33 → £34 → £37
0/12 → 4/14
–
15 Aug
Cook Candles
straight
1
£85 → £80
£37 → £56 → £40
5/12 → 7/14
–
31 Aug
Evergreen (old Tomato vine)
straight
9
£720 → £1,047
£35 → £38 → £41
3/12 → 7/14
12
8 Sep
Evergreen (old Tomato vine)
mixed
20 (2 tested)
£915 → £1,121
£37 → £44 → £45
2/12 → 5/14
10
17 Sep
Evergreen (old Tomato vine)
mixed
27 (5 tested)
£1,064 → £1,444
£45 → £54 → £51
0/12 → 5/14
–
25 Sep
Evergreen (old Tomato vine)
mixed
24 (1 tested)
£1,172 → £1,872
£51 → £49 → £48
3/12 → 3/14
13
4 Oct
Cook Candles
straight
3
£198 → £208
£43 → £35 → £38
7/12 → 2/14
5
4 Oct
Evergreen (old Tomato vine)
straight
44
£1,609 → £2,464
£55 → £44 → £47
4/12 → 2/14
5
12 Oct
Cook Candles
straight
3
£199 → £231
£42 → £38 → £46
3/12 → 3/14
12
12 Oct
Evergreen (old Tomato vine)
straight
8
£2,038 → £2,958
£49 → £50 → £52
4/12 → 2/14
11
6 · What this does and does not tell us
Read with these in mind
Budget changes sit on top of most batches. 14 of 24 batches had spend per day rise 20% or more in the same fortnight. Some of any change in cost per sale is scaling, not the new ads.
Learning phase is not visible after the fact. Meta does not keep learning-status history, so we cannot confirm a campaign re-entered learning each time. We can only see what cost per sale did.
Campaign results blend everything. Old ads keep running underneath. A calm campaign can still be hiding a bad new ad that Meta simply never fed.
Left out of the read (still on record):
17 Apr · Almond April 2026 (3 ads, straight): under £50/day before, too few sales a day to read.
20 Apr · Evergreen 2026 (3 ads, tested): spend fell by more than half: campaign being wound down.
11 Jun · Ocean/Sea Salt (2 ads, tested): under £50/day before, too few sales a day to read.
22 Jul · Seasalt Refresh (Jul 2026) (2 ads, straight): spend fell by more than half: campaign being wound down.
15 Aug · Seasalt Refresh Purchase (Jul 2026) (9 ads, straight): spend fell by more than half: campaign being wound down.
15 Aug · Orangery (Aug 2026) (1 ad, straight): spend fell by more than half: campaign being wound down.
Fir copies of 26 and 27 October: hit by the £5k → £2k budget cut on the same day, and not yet 14 days old.