Round 2 plan and recommendations
Ailis's six-test brief and the 26 new files, reviewed against what the live tests have told us so far.
Copy for every set, ready for your approvals: pott-christmas-round2-copy.pages.dev
1. What is running now, and what to do with it
21 to 28 Oct, Meta-reported, 7-day click. 28 Oct is a part day.
| Test | £/day now | Read | Call | Why |
|---|---|---|---|---|
| Fir Statics campaign | £100 | 1027 £202 / 4 orders / £50 · 1029 £421 / 11 / £38 · 1030 £312 / 9 / £35 | Pause | The winners are known. 1029 is the one Fir copy main is feeding; 1027 is the weakest arm. |
| Fir Videos campaign | £50 | 1033 £766 / 18 / £43 | Keep | Its only funded home. £43 is November break-even. |
| T1 Fir proof | £250 | 1055 £42 · 1056 £37 · 1057 £45 · 1058 £37 (all fed) | Cut to £120 | Answered: the creator demo (1056) is best, all four sit at £37 to £45. Keep as the ring-fence. |
| T2 Fir message | £100 | 1060 £1,233 / 27 / £46 · 1059 and 1061 unfunded for 3 days | Pause | Meta chose the cosy-evening caption and it converts at £46. Nothing more to learn. |
| T3 Starter Pack scent | £300 | 1064 £833 / 30 / £28 (rising) · 1063 £1,443 / 24 / £60 (sliding) · 1062, 1065 unread | Keep | Waits for the Flora four-move agreed on the 28 Oct call. |
| T4 Icicle proof | £200 | 1066, 1067 paused in T4 on 28 Oct · 1069 £165 / 3 on its first fed day · 1068 £16 | Keep ~4 days | Now reading the two ads it never fed. |
2. Why the Fir winners went quiet in main
Christmas ad set spend by day. The lighter block is Ad 1000, still taking about a third of whatever the set gets.
The five Fir copies went live in this set on 27 Oct, the day after the cap. Spend on them since: 1027 £9, 1029 £126, 1030 £7, 1031 £9, 1033 £14. At £701 a day across 24 ads with Ad 1000 and three older ads ahead of them, there is no headroom. The copies are unread, not failed. The lever is the Evergreen budget, not the Fir campaigns.
3. Round 2: six sets, all at once
Every set changes one thing, with its ads competing inside it, one caption, built paused. Budgets below total £450 a day across 23 ads, which is thin: a clean 7-day read wants about £150 per ad. Expect Meta to fund roughly half the arms in week one. An unfunded arm is unread, not a loser, and gets its turn when T3 and T4 roll their £275 into these sets. If anything has to give later in the week, it gives in this order: Ailis to camera first, then Noël and Starlight, then Refilling; the review statics and the PR pair last, because they shape November creative.
Christmas review statics: which line sells?
Launch nowThe question. Which customer line on a still sells: true to life, not fake, fills the room, limited edition, or the pottery?
How it reads. One format throughout, the review card that already works, so the line is the only variable. Five Fir stills and two Starlight. Scent is a known confound: read the Starlight pair against 1066 (the Starlight still already running at £33).
Copy. Ad 1000's live copy pool on every arm (it names all three scents), nothing new to approve.
Watch out. The "Smells like a fir tree, gorgeous!" still is left out: same line as 1027 on a more polished photo, and the Fir test already showed polish does not change the result. Typos "ofr", "cinamon", "begamot". All seven quotes must be genuine reviews.
Noël: which angle makes the scent compelling
Launch nowThe question. Does Noël sell on a personal recommendation, the look of the Pott, or the story behind the scent?
How it reads. Same caption on all arms. All three name cinnamon and clove. Read on cost per order once each arm has £150, with 1064 (sniff plus a named scent, £28) as the benchmark.
Copy. New Noël caption, two options on the review page.
Watch out. The Holly gift-box video names no scent and opens on the present, the frame that got clicks not buyers on 1063. It goes to Gifting instead.
Starlight: who tells the "not too Christmassy" story
Launch nowThe question. Does the non-Christmassy winter idea land better from Ailis, from a creator concept, or from an unboxing?
How it reads. Videos only, so format is constant. The two Starlight stills moved to set 1 with the other statics. Read against 1067 (kiln video, £30) which shares the caption.
Copy. Icicle 02 caption (approved), same as T4, so the arms read against 1066 and 1067.
Watch out. Dropped the 8-second review-on-the-Grand video: the Grand photo lost in the Fir test (1028, £71) and its review line contradicts the still. Starlight notes disagree across the set (see flags).
Refilling: which explanation converts
Launch nowThe question. Does understanding the refill remove hesitation, and which explanation does it best: a customer question answered, a full walkthrough, or a close-up demo?
How it reads. Ailis's test as she wrote it. Speaker is a known confound (two Ailis, one creator). Read add-to-basket per landing-page view as well as cost per order, since the point is removing hesitation.
Copy. New Starter Pack caption (how-to, not ritual: the videos are practical), two options on the review page.
Watch out. The close-up is Ad 1042's video, already live to the zero-order email list; cold is new for it. All three say "about 30 seconds" (claim register). The walkthrough ends on a pink production slate at 1:07 that must be trimmed.
PR: does press proof improve conversion
Launch nowThe question. Does a newspaper masthead on the image add enough credibility to pay for itself?
How it reads. One concept three ways: a press quote and masthead over a product photo. Ailis supplied the press side only. The clean control is the same three images with the masthead removed and the quote left in. She needs to export those. Meta's A/B tool splits the budget evenly between the two ad sets, so each side gets the same chance.
Copy. New caption with no proof of its own, so the masthead is the only difference.
Watch out. The Telegraph video is a different concept (Ailis talking through the article) with no like-for-like pair, and an A/B test cannot hold an extra ad on one side without changing what that side is. It runs in set 6 with the other Ailis videos instead.
Ailis to camera: which topic earns more creative
Launch nowThe question. With Ailis on camera every time, which subject sells: sizes, the MatchPott, the founder story, or the Telegraph piece?
How it reads. One speaker, one format, so the topic is the only variable. This is the set that writes next week's shoot list. An unfunded arm counts as a result.
Copy. New Starter Pack caption that sits on all four videos, two options on the review page.
Watch out. The Telegraph video sits here, not in the PR test: it is Ailis to camera like the other three, and an A/B test cannot carry an odd ad out. In it she says "Diptyque uses paraffin" while the article on screen says a vegetable and paraffin blend, and says six years in business where the market video says four. Burn times in the sizes video match the website.
Outside the tests
4. Where the winners go
The five sets already split by destination. Review statics, Noël and Starlight winners go into the Christmas ad set, and they test on its settings and landing page. Refilling, PR and Ailis-to-camera winners go into Everything Else (the PR images show the Ocean Pott, and those videos are not seasonal), and they test on the Starter Pack page. The 28 Oct analysis showed the same ad can do £23 in one main ad set and £57 in another, so matching the test to its destination is worth keeping as a check, even though it does not change the structure.
5. Budget picture: testing held at 20%
| Block | £/day | Notes |
|---|---|---|
| Evergreen (capped) | £2,000 | The lever if testing needs more room: every £100 here allows £25 more testing. |
| Gifting | £400 | 1063 now takes about 80% of it. |
| Cook, retention, subscription | £322 | Unchanged. |
| Round 1 ring-fences: T3 £100 (1064), T1 £75 (1056), T4 £100 (1068, 1069) | £275 | Insurance while their Christmas-set copies prove they get delivery. T4 reads in about four days. Each one pauses once its copy in main is spending, and the money rolls into round 2. |
| Round 2: statics £100, PR £80, Refilling £75, Ailis to camera £75, Noël £60, Starlight £60 | £450 | Six ad sets, 23 ads, all live at once. |
| Testing total | £725 | 21% of £3,447 for about four days, then T4's £100 comes off and it is 18.7%. Fir Videos paused. |
Round 1 ad sets have been running at £28 to £46 per order against a November break-even of about £43, so the 20% is not a drag on the account. The constraint is the read, not the money: £450 across 23 ads is about £135 per ad over a week if Meta spread it evenly, which it will not. Next week's decisions should treat anything under about £75 as unread.
6. Flags for Ailis before anything is built
The 68-second video ends on a pink card reading "Brand Video, Refill into Pott". It needs trimming before upload.
The claim register bans timed-refill claims until timed on camera. Ad 1042 already runs it, so this is Casey's call. Our captions make no timed claim.
The scent static says "not cinnamon". Ailis's video says bergamot, tonka and cinnamon. The website says bergamot, lemon, mandarin and cinnamon. One version has to be chosen.
The PR video says Pott started six years ago. The founder-at-market video says four.
The Telegraph screenshot in the same video says a vegetable and paraffin blend. Safer to cut the line.
The test needs the same three images without the masthead. Ailis has to export them.
"ofr yourself" (Fir 4), "cinamon" and "begamot" (Starlight scent static). Customer quotes can stay as written if they are genuine; worth asking.
Six new customer quotes. We never run one that is not a real review.
7. What happens next
- Casey sets T3 to £100, T1 to £75, T4 to £100 and pauses Fir Videos.
- Ailis approves copy on the review page and answers the four flags.
- All six ad sets built paused as Ads 1071 onward, screenshots sent, then switched on. The PR set waits only for the no-masthead exports.
- Next week: read every arm, treat under £75 as unread, decide what to shoot next and roll T3 and T4 budget into the funded winners.
- Reads per ad at £150 (winner at or under £40 with 6 or more orders, kill over £75) and per ad set around day 7.
Related pages: Fir test results · Does testing predict main? · What we're testing