Pott Candles - Content Diversity Tracker

Who we are reaching, and at what stage

Inferred, not recorded. Our ad names have no slot for persona or awareness stage, so these come from what each ad ARGUES, through a stated mapping in the client config with a reason on every rule, sourced to the persona doc. It is the weakest source we have: a tag written at launch would replace it outright. Read it as "this is who this creative is written at", not as who Meta delivered it to.

	Problem-aware	Solution-aware	Product-aware	Most aware	Total
The Scent-Led Home Curatorcore self-buyer	3.0%	60.3%	29.5%		92.7%
The Refill Loyalistmost-aware, the economics persona				5.4%	7.3%
The Gift-Giverthe Q4 engine					0.0%

No creative at all for: Unaware. Coverage is a creative question, never a delivery one: the fix is to make that ad, never to build an audience or an ad set for it.

What kind of ad each persona actually gets

The Scent-Led Home Curator

Proof 34.0% · Showcase 25.6% · To camera 16.1% · Unboxing 15.0%

The Refill Loyalist

Showcase 45.2% · Teaching 26.2% · To camera 8.7% · Unboxing 7.9%

The Gift-Giver

Showcase 100.0%

What we claim × how we show it

Down the side is what an ad CLAIMS, the one thing it leads on, read from the primary text rather than from the ad name. Across the top is what it DOES, the move it makes on the viewer. Darker is more spend, and the edges carry each row and column total.

This table answers one question: are we making every claim, and are we stuck making it one way? A row that is all in one column is a claim with a single execution behind it. The table below it uses the same rows to ask a different question.

↔ scroll for the rest of the columns

	Proof 32.0%	Showcase 27.0%	To camera 15.6%	Unboxing 14.5%	Slice of life 5.4%	Teaching 4.5%	All
RefillableP1 you keep the Pott, the candle refills it			3.2%				6.2%
ValueP1 what it costs you over time							0
Worth keepingP2 the object is good enough to live with			3.4%				4.4%
Artisan madeP2 hand thrown, named makers, British	24.1%	4.9%	4.1%	12.4%	2.4%		49.8%
RitualP2 how it fits your evening							0
The scentP3 the notes and the throw	2.2%	10.1%	3.6%			2.7%	18.9%
Seasonal dropP3 new, and of this season		5.5%					7.3%
SubscriptionP3 the rotation, automated		3.3%					7.1%
How it worksP3 the mechanic, explained							0.9%
A job it doesP3 cooking smells, citronella	2.1%						5.4%
GiftingP1+2 bought for someone else							0

Not running at all: Value · Ritual · Gifting. A whole row empty is a bigger finding than any single empty cell, because the account is not making that argument in any form at all. But over the last 365 days Ritual ran £2,184 at 3.14x, and Gifting ran £55,146 at 2.01x, so those are switched off rather than never tried. A 30-day window cannot tell those two apart, which is why what already works reads twelve months. The genuinely untried one is Value.

Read the rows before the cells

Artisan made P2

49.8% of spend · 80 ads · £223 per ad
last new ad 2026-09-13

The scent P3

18.9% of spend · 94 ads · £72 per ad
last new ad 2026-08-15

Seasonal drop P3

7.3% of spend · 28 ads · £94 per ad
last new ad 2026-08-31

Subscription P3

7.1% of spend · 19 ads · £134 per ad
last new ad 2026-07-02

Refillable P1

6.2% of spend · 132 ads · £17 per ad
last new ad 2026-09-10

A job it does P3

5.4% of spend · 45 ads · £43 per ad
last new ad 2026-08-15

Worth keeping P2

4.4% of spend · 15 ads · £104 per ad
last new ad 2026-09-10

How it works P3

0.9% of spend · 3 ads · £112 per ad
last new ad 2026-09-10

Value P1

nothing

Ritual P2

nothing

Gifting P1+2

nothing

What we claim × how we prove it

Same rows, different question: why should she believe it? Across the top is the evidence behind the claim. A customer saying it, the rating, a press clipping, a demonstration, a comparison against what she would otherwise buy, the maker's own word, or a checkable fact the claim carries with it.

The last column is the one to read first. Asserted only means the ad states the claim and offers nothing to support it. It is a real column, not a tagging failure, and on most accounts it is bigger than anyone expects.

↔ scroll for the rest of the columns

	Customer says it 20.6%	Reviews or rating 32.3%	Shown working 4.5%	Against the alternative 7.9%	Maker says it 13.0%	Carries its own detail 9.2%	Asserted only 12.5%	All
RefillableP1 you keep the Pott, the candle refills it					3.4%			6.2%
ValueP1 what it costs you over time								0
Worth keepingP2 the object is good enough to live with	3.4%							4.4%
Artisan madeP2 hand thrown, named makers, British	11.5%	24.1%			5.2%	7.1%		49.8%
RitualP2 how it fits your evening								0
The scentP3 the notes and the throw		2.2%	2.7%	6.0%	3.6%	2.1%	2.2%	18.9%
Seasonal dropP3 new, and of this season							5.2%	7.3%
SubscriptionP3 the rotation, automated	4.1%							7.1%
How it worksP3 the mechanic, explained								0.9%
A job it doesP3 cooking smells, citronella		2.1%					2.4%	5.4%
GiftingP1+2 bought for someone else								0

How this is read. Proof comes from the creative wherever the creative settles it: if a customer carries the ad it is a testimonial, if the ad IS a review card it is the rating, if it shows the thing happening it is a demonstration. Only where neither the messenger nor the form answers does the copy decide. That means some ads give the same answer to both tables by construction, and those cells are the least interesting ones here. The cells worth reading are where the two tables disagree, for example a claim about the makers carried by a customer rather than by the maker.

Never used as proof: Press or editorial. Worth asking whether that is a decision or an oversight.

Where the evidence sits

Reviews or rating

32.3% of spend · 87 ads

A customer says it

20.6% of spend · 72 ads

The maker says it

13.0% of spend · 51 ads

Asserted only

12.5% of spend · 82 ads

Carries its own detail

9.2% of spend · 39 ads

Against the alternative

7.9% of spend · 50 ads

Shown working

4.5% of spend · 36 ads

What our ads actually do

The move each ad makes on the viewer: talks to them, shows them, teaches them, opens a box in front of them. Not what it argues and not how it looks, just what it does. Tap any i for what a lane counts as and, more usefully, what it does not. Spend-weighted, last 30 days. A lane can hold plenty of budget while every ad in it is months old, so the ageing badge matters as much as the bar.

Proof is the creative

32.0%86 ads

Product presented, no story

27.0%107 ads

Person speaking to camera

15.6%78 ads

Unboxing, the reveal

14.5%27 ads

A slice of life, no address

5.4%59 ads

Teaching, steps, how-to

4.5%35 ads

Comparison, before & afterageing 90d

1.0%1 ad

Not stated

0.0%22 ads

Shows it workingstale 151d

0.0%1 ad

Enumerated, text-ledstale 130d

0.0%1 ad

Empty by choice: Reacting, replying · Skit, humour, a bit · A story with beats. Parked for this brand with a reason on file, so they are not in the build list. Parking matters as much as the gap: it stops the report reading as a demand for twelve of everything.

What moved since 2026-09-09

Bars sit either side of the zero rule, so direction is position, not colour. A lane shrinking is not automatically bad. The two flags mark the cases worth a decision.

Proof is the creative

+2.995 → 86 ads

Unboxing, the reveal

-0.929 → 27 ads

Teaching, steps, how-to

-0.636 → 35 ads

Person speaking to camera

-0.599 → 78 ads

What has worked, and what has not

Every ad that spent in the last 365 days. Start with the two summaries: one row per form and one row per claim, each with the number of ads we made, what they spent, what they returned and whether anything from them is running now. The return is shown against the account's 1.90x average, and dimmed where it came in under it. Click any row for the ads inside it.

By form: what the ad does

↔ scroll for the rest of the columns

Form	Ads	365d spend	Share	Return	Per ad	Running now
Showcase	237	£154,972	39.6%	1.95x	£654	£8,502
Teaching	133	£55,367	14.1%	1.90x	£416	£1,601
Unboxing	84	£51,612	13.2%	1.77x	£614	£5,182
To camera	170	£49,144	12.5%	1.71x	£289	£6,752
Slice of life	191	£43,447	11.1%	2.10x	£227	£1,940
Proof	181	£35,811	9.1%	1.76x	£198	£11,455
Compare	4	£814	0.2%	2.58x	£204	£370
List	3	£318	0.1%	1.14x	£106	off
Demo	5	£311	0.1%	0.68x	£62	£4

By claim: what the ad says

↔ scroll for the rest of the columns

Claim	Ads	365d spend	Share	Return	Per ad	Running now
Seasonal drop	146	£105,076	25.4%	1.91x	£720	£2,629
Refillable	304	£64,805	15.6%	1.67x	£213	£2,224
Gifting	46	£55,146	13.3%	2.01x	£1,199	off
Artisan made	218	£54,562	13.2%	2.09x	£250	£17,830
The scent	147	£39,793	9.6%	1.98x	£271	£6,769
A job it does	65	£38,636	9.3%	1.65x	£594	£1,921
Worth keeping	95	£33,637	8.1%	1.81x	£354	£1,564
Subscription	44	£15,922	3.8%	2.59x	£362	£2,546
How it works	6	£4,579	1.1%	0.80x	£763	£337
Ritual	4	£2,184	0.5%	3.14x	£546	off

The detail: each claim and form together

How often did we make this, and what happened when we did? Two numbers decide it, and both are this account measured against itself. Funding is spend per attempt against the £385 the average ad in this account received: Meta deciding, with real money, whether to keep feeding a thing we made. Return is the recipe's ROAS against the account's 1.90x. Fewer than 4 attempts, or less than £898 spent, gets no verdict, because silence is not evidence and four ads sharing £80 can show any return at all. Each band shows its biggest 6; the strip above counts all of them.

Worked

125 ads

£121,514 · 12 recipes

Funded above average and returned above it. Make more of these.

Sold, not efficiently

157 ads

£134,080 · 12 recipes

Meta kept feeding it, the return lagged. A winner to fix, not a failure.

Under-funded

370 ads

£74,893 · 13 recipes

Returned above average and never got scaled. Put money behind it before making more.

Not working

300 ads

£53,463 · 9 recipes

Made repeatedly, funded below average, returned below average. Stop.

Too few to call

123 ads

£30,389 · 38 recipes

Fewer attempts than the floor. No verdict either way.

↔ scroll for the rest of the columns

Verdict	For	Claim	Form	Times made	Spend	Per ad	Return
Worked	P3	Gifting	Showcase	15	£26,575	£1,772	2.17x
Worked	P1	Artisan made	Unboxing	9	£17,568	£1,952	1.97x
Worked	P3	Seasonal drop	Showcase	6	£16,087	£2,681	1.95x
Worked	P1	Gifting	Showcase	5	£11,159	£2,232	2.02x
Worked	P1	Seasonal drop	Slice of life	26	£10,882	£419	2.62x
Worked	P1	The scent	Teaching	17	£8,946	£526	2.29x
Sold, not efficiently	P1	Seasonal drop	Showcase	38	£53,433	£1,406	1.62x
Sold, not efficiently	P2	Refillable	Teaching	15	£14,409	£961	1.47x
Sold, not efficiently	P1	Worth keeping	To camera	17	£13,437	£790	1.20x
Sold, not efficiently	P1	A job it does	Slice of life	19	£13,306	£700	1.58x
Sold, not efficiently	P1	Refillable	Unboxing	28	£11,323	£404	1.26x
Sold, not efficiently	P1	A job it does	Not stated	8	£7,816	£977	1.44x
Under-funded	P1	The scent	Showcase	63	£15,794	£251	2.04x
Under-funded	P1	Artisan made	Showcase	32	£10,352	£323	1.96x
Under-funded	P1	Worth keeping	Not stated	28	£8,888	£317	2.81x
Under-funded	P1	Refillable	Showcase	29	£7,204	£248	2.44x
Under-funded	P1	Refillable	Slice of life	35	£6,948	£199	1.95x
Under-funded	P1	Artisan made	Slice of life	61	£6,429	£105	2.59x
Not working	P1	Refillable	To camera	82	£15,217	£186	1.74x
Not working	P1	Artisan made	Proof	32	£9,699	£303	1.74x
Not working	P1	Seasonal drop	Proof	23	£7,883	£343	1.81x
Not working	P1	Refillable	Not stated	41	£5,053	£123	1.53x
Not working	UNASSIGNED	Worth keeping	Teaching	17	£4,910	£289	1.52x
Not working	P1	The scent	Proof	14	£4,322	£309	1.52x
Too few to call	P1	Gifting	Unboxing	2	£7,519	£3,759	1.67x
Too few to call	P1	How it works	Teaching	1	£4,084	£4,084	0.63x
Too few to call	P3	Seasonal drop	Teaching	3	£3,196	£1,065	2.18x

Read the bottom two rows of the grid together. 300 ads sit in recipes that were funded below average AND returned below it, and 370 ads sit in recipes that returned ABOVE average and still never got funded. Those need opposite decisions: stop making the first, put money behind the second before making any more of it. ⚠️ Spend per ad is partly a function of which campaign an ad sat in, so a small budget can look like a verdict. Across a hundred attempts that is a pattern. Across four it may be an artefact, which is why four is the floor. The claim could not be read on 7.8% of 365-day spend, mostly sale creative that makes no claim at all and older ads with no body text. Those are excluded from the benchmarks rather than counted as a content type.

What already works

The same 365 days, asked about right now: of what we have made, what is running, what is tiring and what is switched off. Every ad down to the top 100.0% of spend, rolled up into RECIPES: who it is for, what it claims, what it does. The recipe is the unit on purpose. When an ad tires you do not rebuild that ad, you rebuild its recipe on a different scent or a different product, and that is a brief a person can act on. 654 of these 1266 ads carry judgements from the weekly register, including its vision passes and hand corrections; the other 612 were tagged fresh through the same rules. Ranked on spend. Click a row for the ads inside it.

↔ scroll for the rest of the columns

State	For	Claim	Form	Proof	365d spend	Return	Ads	Newest
FRESH	P1	Seasonal drop	Showcase	Customer says it	£53,433	1.62x	38	15d
SEASONAL	P3	Gifting	Showcase	Asserted only	£26,575	2.17x	15	308d
FRESH	P1	Artisan made	Unboxing	Customer says it	£17,568	1.97x	9	31d
FRESH	P1	unclassified	Not stated	Asserted only	£16,946	3.54x	61	32d
SEASONAL	P3	Seasonal drop	Showcase	Asserted only	£16,087	1.95x	6	320d
FRESH	P1	The scent	Showcase	Asserted only	£15,794	2.04x	63	32d
FRESH	P1	Refillable	To camera	Maker says it	£15,217	1.74x	82	52d
STALE	P2	Refillable	Teaching	Shown working	£14,409	1.47x	15	74d
FRESH	P1	Worth keeping	To camera	Customer says it	£13,437	1.20x	17	5d
FRESH	P1	A job it does	Slice of life	Asserted only	£13,306	1.58x	19	31d
FRESH	P1	Refillable	Unboxing	Customer says it	£11,323	1.26x	28	5d
SEASONAL	P1	Gifting	Showcase	Asserted only	£11,159	2.02x	5	287d
FRESH	P1	Seasonal drop	Slice of life	Asserted only	£10,882	2.62x	26	15d
FRESH	P1	Artisan made	Showcase	Carries its own detail	£10,352	1.96x	32	2d
FRESH	P1	Artisan made	Proof	Reviews or rating	£9,699	1.74x	32	2d
STALE	P1	The scent	Teaching	Shown working	£8,946	2.29x	17	83d

State is read from the data, not judged. FRESH is running and fed inside 60 days. STALE is running on an asset older than that. OFF has spent nothing in 30 days. SEASONAL is off because every asset in it launched in October, November or December, which is a season and not neglect.

Where the biggest gaps are

Ranked by what it costs us to leave it open, not by how empty the cell is. The grids above can only see arguments we already make; this is the read on the ones we do not, and on the one we make too often.

1

We keep making the same recipe in batches, and Meta funds one ad per batch

12-monthwhat has not worked

Refillable carried by a person speaking to camera is the most-attempted recipe in the account: 82 launches of 52 different ads in 12 months, £15,217 at 1.74x against the 1.90x average, and £186 per launch against £385. The money went to a handful of single ads: Ad 344 in February (£7,039 at 1.61x) and Ads 503 and 504 in June (£4,591 at 1.84x). Everything since has gone out in batches, and every batch ended the same way.

Evidence. 12-month ad-level pull. An earlier version of this page said 143 attempts. 61 of those were customer reviews typed over a product photo, which are stills, not a person on camera. They are corrected and now sit under proof is the creative, and the fourth card below is that batch.

Ad 648 | Video | Citronella | Founder Story | Founder VO | None | OG

14 July: Citronella founder story, Ads 637-652

16 ads · £2,672 · 1.94x

Meta funded Ad 648 (£1,877 at 2.10x). The other 15 shared £795, and 11 never made a sale.

Ad 623 | Video | Seasalt | Founder Story | Founder VO | None | OG

3 July: Seasalt founder story, Ads 617-626

10 ads · £221 between them

No ad in the batch reached £75. Six never made a sale.

Ad 765 | Video | New Customer Event | UGC Testimonial | UGC | None | OG

25 July: customer videos, Ads 731-771

21 ads · £129 between them

The biggest took £40. Seventeen never made a sale. Too little spend to read a return.

Ad 725 | Image | Citronella | UGC Testimonial | UGC | None | OG

25 July: the same claim as review stills, Ads 714-757

37 ads · £1,764 · 1.59x

Meta funded Ad 725 (£1,204). Thirty never made a sale.

So what. Volume is not diversity. A batch of ten versions of one recipe bought one ad’s worth of learning each time, because Meta picked one and starved the rest. Stop adding variants here and put the effort into the under-funded recipes that returned above average.

2

Gifting is ready to rebuild for November: £55,146 at 2.01x last Q4

12-monthseasonal: November

Gifting shows zero spend on the last 30 days, and that is correct: it is a Q4 lane and this is September. Across the last 12 months it ran 46 ads for £55,146 at 2.01x, better than the 1.90x account average. The biggest single recipe is the gift set presented as a still, £26,575 at 2.17x across 15 ads.

Evidence. 12-month ad-level pull, 2025-09-16 to 2026-09-15. Ads 213, 251, 194, 192, 217, 214, 189, 257, 245, 263. Every one launched between 30 Oct and 24 Dec 2025, and the first went live on 11 Nov.

So what. Not a launch for now, a build for November. We already have last year’s winners, their copy and their formats, so rebuild them rather than invent new gifting creative, and have it ready before the gifting campaign opens on 1 Nov.

3

Value is the real empty row, and 12 months confirms it

confirmed on 12 months

Value has zero spend across the whole 12 months. The approved value line appears in copy worth £3,686, which is 0.8% of everything spent, and it has never led an ad.

Evidence. Search of all 1,266 primary texts for the locked line and its variants. Price is 60% of poll hesitations and 26% of the 168 low-star reviews.

So what. This is the one genuinely untested claim with a business case behind it. It is also the cheapest to test, because the line is already written and already approved.

4

Our one refill how-to video lost money when it opened the ad

12-monthone data point

Ad 353 was a How To Refill video: take the old wax insert out, drop a new candle refill in, light it. It ran in the evergreen campaign from 20 Feb, spent £4,084 and brought back 63p for every £1 spent, against a 1.90x account average. Five small statics that also explain the refill add £495, so almost all of the evidence is this one video.

Evidence. 12-month pull. How it works: 6 ads, £4,579 at 0.80x, of which Ad 353 is £4,084. Meanwhile 11% of the 168 low-star reviews are about getting the old wax out, so the need is real.

So what. One video failing is one data point, not proof that showing the refill cannot sell. The safer test is to show the refill as support inside an ad that opens on something else (the scent, the pot, the price), not as the opening.

5

Ritual ran four times, at the best return in the account

corrects the 30-day read

Ritual reads as an empty row on 30 days. Over 12 months it ran 4 ads for £2,184 at 3.14x, the highest return of any claim. The body leads on the sound of the match and the evening, not the object or the scent.

Evidence. Ad 132 and its variant, launched 2025. 12-month pull.

So what. Small sample, so it earns a proper test rather than a queue. But it is a revive, not an invention, and the copy that did it already exists.

6

Press is empty on 30 days and on 12 months

confirmed on 12 months

No ad in the last year has used a press or editorial mention as its proof. The Ideal Home placement has been held since February and has still never run.

Evidence. Proof axis, both windows.

So what. A graphic, not a shoot. It is the only proof column with nothing in it at all.

What to make next

Twelve months of spend says the account already knows what works. Gifting ran £55,146 at 2.01x last Q4 and none of it is running yet. Ritual returned 3.14x on four ads and was never made again. Value has never led an ad at all. Meanwhile most of this summer’s production went into batches of one recipe, refillable told to camera, where Meta funded one ad per batch and starved the rest. The list below is ordered by what each move costs.

REBUILD

Attach the proof to the dropP1

£1,860 of drop spend is running right now with no evidence attached. Re-edit three statics, attach the rating and one verified line. No shoot, no new design.

never made
REVIVE

Subscription, showcaseP2

£6,985 over 365 days at 2.72x. Nothing from it has spent in 30 days. Newest asset 439d old.

£6,985 2.72x
REVIVE

Refillable, teachingP1

£2,520 over 365 days at 1.81x, under the 1.90x average. Nothing from it has spent in 30 days. Newest asset 165d old.

£2,520 1.81x
REVIVE

Ritual, teachingP2

£2,184 over 365 days at 3.14x. Nothing from it has spent in 30 days. Newest asset 361d old.

£2,184 3.14x
REVIVE

Seasonal drop, to cameraP1

£882 over 365 days at 1.35x, under the 1.90x average. Nothing from it has spent in 30 days. Newest asset 327d old.

£882 1.35x
REBUILD

The scent, teachingP1

£8,946 over 365 days at 2.29x, £900 still running. Newest asset 83d old.

£8,946 2.29x
REBUILD

Subscription, to cameraP2

£6,399 over 365 days at 2.08x, £1,406 still running. Newest asset 75d old.

£6,399 2.08x
REBUILD

Seasonal drop, unboxingP1

£5,197 over 365 days at 1.94x, £129 still running. Newest asset 334d old.

£5,197 1.94x
REBUILD

A job it does, unboxingP1

£4,236 over 365 days at 1.72x, under the 1.90x average, £165 still running. Newest asset 119d old.

£4,236 1.72x
NEW

Say the price out loudP1

Fills an empty row using a line that is already written and already approved.

never made

Prepare for Q4: last year's proven seasonal work

These are proven, but they are seasonal. They belong in the Q4 build, not in this month's launches. The date is when last year's version went live, which is the latest the rebuild should be ready by.

PREPARE

Gifting, showcaseP3

£26,575 over 365 days at 2.17x, all of it last Q4. Last year it went live on 11 Nov. Out of season now, so this is prep for the season, not a switch-on.

£26,575 2.17xlive 11 Nov last year
PREPARE

Seasonal drop, showcaseP3

£16,087 over 365 days at 1.95x, all of it last Q4. Last year it went live on 16 Oct. Out of season now, so this is prep for the season, not a switch-on.

£16,087 1.95xlive 16 Oct last year
PREPARE

Gifting, unboxingP1

£7,519 over 365 days at 1.67x, under the 1.90x average, all of it last Q4. Last year it went live on 11 Nov. Out of season now, so this is prep for the season, not a switch-on.

£7,519 1.67xlive 11 Nov last year

Four moves. The first three are ordered by what they cost. Revive is proven and switched off, so it costs nothing but a decision. Rebuild is proven, running and tiring, so it is the same recipe on a different scent or product, never a rebuild of the ad itself. New is the only kind that needs a shoot day. Prepare is proven seasonal work that is out of season, so it is rebuilt ahead of its date rather than switched on now. Revive and rebuild are read off 365 days of ad-level spend, which averaged 1.90x; ranked on spend, with the return shown beside it and flagged where it came in under that average. Click any row to see the ads inside it.

£449,202

365-day spend read

top 1266 ads, 100.0% of everything spent

£35,822

running now

417 ads live

12.5%

claimed with no proof at all

asserted only, see the second grid

3

claims not running

Value, Ritual, Gifting

What needs deciding

We keep making the same recipe in batches, and Meta funds one ad per batch

Volume is not diversity. A batch of ten versions of one recipe bought one ad’s worth of learning each time, because Meta picked one and starved the rest. Stop adding variants here and put the effort into the under-funded recipes that returned above average.
Gifting is ready to rebuild for November: £55,146 at 2.01x last Q4

Not a launch for now, a build for November. We already have last year’s winners, their copy and their formats, so rebuild them rather than invent new gifting creative, and have it ready before the gifting campaign opens on 1 Nov.
Value is the real empty row, and 12 months confirms it

This is the one genuinely untested claim with a business case behind it. It is also the cheapest to test, because the line is already written and already approved.

Each of these is argued in full in where the biggest gaps are, and the evidence for every row in the list is in what already works.

What to build next

Ordered by what each asset costs, not by how big its gap is. The revives at the top need a decision and an export, not a shoot. The gifting work is a rebuild of last Q4's winners rather than new creative, so the copy and the formats already exist. Only the value row and the press card are genuinely new. Capped for the six weeks before Q4 trading, and every row names a CLAIM and the FORM or PROOF it is missing, because one gifting still closes a claim row and a form lane at once. Absent lanes get two to four shots, not one: at a roughly 5% winner rate a single ad cannot tell you whether the lane failed or the execution did.

#	Cell	What to make	Why now
1	How it works x demonstrationP1	Wax out, candle refill in, as proof under a claim that sellsOne continuous phone shot, no voiceover. Warm water, the old wax disc lifts out clean, the new candle refill drops in, the lid goes on. Thirty seconds. The change from Ad 353 is the FRAME: this runs underneath a claim that already sells, never as a how-to hook. Do not open on the steps. Do not claim a time on screen until someone has timed it.	Ad 353 ran the steps as the hook at £4,084 and returned 0.63x, so the idea is not untested, the execution was wrong. The demand is real: 11% of 168 low-star reviews are about getting the old wax out.
2	Refillable x demonstrationP1	The swap, on a scent that is not CitronellaSame mechanic, shot on Orangery or Seasalt. The point is to move the refillable claim off a summer SKU before the season ends. House range only.	90.5% of refillable spend is Citronella. The claim needs a home that survives October.
3	Value x person speaking to cameraP1	Say the price out loudStraight to camera. You buy the Pott once. After that you are only ever buying the candle. Name what she keeps, then what it costs. Value before price, benefit framing only, no landfill statistic and no sustainability guilt.	Fills an empty row using a line that is already written and already approved.
4	Value x comparisonP1	One Pott beside the jars it replacesSide by side, priced. One Pott against the stack of jars the same money would have bought and binned. What she saves and what she keeps. No guilt angle.	Fills an empty row and a nearly empty form in one shoot. Comparison holds one ad, 90 days old.
5	Value x proof is the creativeP2	The ladder, as mathsA designed card. First purchase, then each candle refill, then the same again next season. Show the line flattening. Numbers only, no adjectives.	Value is unproven as well as unsaid. A number is its own evidence.
6	Gifting x unboxingP3	The recipient opens it, rebuilt off last Q4Shot from the receiver's side. Hands, paper, the Pott appearing. Last year's gifting unboxings (Ads 217 and 214) ran £7,519 at 1.67x, below the £26,232 at 2.17x the presented gift set did, so treat unboxing as the second format, not the first. No price on screen.	Proven format, switched off. Rebuild rather than invent, and lead with the stronger of the two.
7	Gifting x person speaking to cameraP3	The moment you think of them, written for himTo camera, plain language, written for a man buying for a woman. No scent notes, no fragrance vocabulary. Nothing in last Q4's gifting creative was written to him, so this is the one genuinely new angle inside a proven claim.	The December buyer goes from 6.8% male to 20.3%, and none of the 15 gifting ads addressed him.
8	Gifting x product presentedP3	The gift set, presented. The single best recipe we haveRebuild the recipe that earned £26,232 at 2.17x: the set presented as a styled still, contents stated, gift framing in the copy. State what is in the box. Pott, first candle refill, MatchPott, £69. Last year's line was: artisan made, endlessly refillable, beautifully thoughtful, the kind of gift that becomes part of your home.	Highest-return gifting recipe in the account and nothing from it is running. This is the first asset to make.
9	Seasonal drop x reviews or ratingP1	Attach the proof to the dropRe-edit, not a shoot. Take the three best-performing drop statics and add the rating and one verified customer line about that scent. Verify every quote against Yotpo before it runs.	£1,860 of drop spend is asserted with nothing behind it. This is the cheapest proof fix in the account.
10	Artisan made x house rangeP1	The making, off Tomato VineThe maker story shot on the house range. Wheel, glaze, stamp, the same craft argument that already works, with no collaboration attached.	96.2% of the maker story is one product. The pillar is untested on its own range.
11	Ritual x a slice of lifeP1	The evening it belongs to, rebuiltNo dialogue. The Pott lit at the point in the evening it actually gets lit. Last year's version opened on the sound of the match. One test, not a queue.	Ritual ran £2,071 at 3.17x, the best return of any claim, on two ads. Small sample, real signal.

Method

Meta ads with 30-day spend in evergreen trading. Pott runs no sale campaigns (see the urgency-offer PARK), so ALL campaigns are evergreen trading and none are excluded. If a sale campaign is ever launched, add excludeCampaign.
Six axes, all capped: persona 5, stage 5, what-it-claims 11, messenger 8, what-it-does 12, look 6. A new value needs a merge or a retirement, enforced in code. Specificity comes from combinations, and anything that still wants its own bucket becomes a free topic tag that is never scored. What an ad CLAIMS is client-specific and is not ours to invent: the territories come from clients/pott-candles/brand-messaging.md sections 1 and 5b, DRAFT 16 Sept 2026. Pillars set by Casey. Territories re-cut 16 Sept on Casey's direction: the eight brand territories were split into eleven CLAIMS, each named the way an ad would say it rather than as an abstraction, so a row can be handed to a copywriter as a brief. Proof mechanisms (social proof, us-vs-them, objection handling, press) are NOT here: they are the separate proof axis. Rules 33-71 added 17 Sept 2026 from the 12-month top-spend read (44 distinct bodies, Q4 2025 onward). APPENDED so they cannot outrank the 32 rules the live 15 Sept report is tagged against.
Judgements are keyed by ad id and carried forward. A weekly run only tags ads it has not seen, and never overwrites a tag written at launch, corrected by hand, or bought with a vision pass.
Register: 654 ads. Snapshots held: 3 (2026-08-18 to 2026-09-15).
This allocates production, never delivery. A gap is filled by making that creative, never by building an audience or an ad set for it.
What we still cannot read. Our ad names say who and how reliably; they say what the ad DOES about half the time, because the naming convention never had a slot for it. Unlabelled right now: 0.0% of spend on what-it-does (22 ads). Tagging at launch is the fix; a one-off vision pass clears the back catalogue.
What an ad CLAIMS is read from its primary text, the words we chose on purpose, except where an ad name states the claim more precisely than a shared body does. The four Seasalt hook tests are the case that proves the rule: one body across four videos, so only the launch name separates them. Where a spoken hook differs from the primary text, this reads the text.
Persona and awareness stage are inferred from the argument for 100.0% of spend and left blank for the rest. Inference says who the creative is written at, which is a reasonable proxy but not the same as someone deciding at launch.
Parked by choice:
- Reacting, replying - Off-brand. Reaction and duet formats fight the calm, crafted register the brand sells on.
- Skit, humour, a bit - Off-brand. The meme and skit register undercuts a premium hand-thrown ceramic product.
- A story with beats - Deprioritised. Hero brand films are expensive and rarely earn direct-response delivery at this spend level.

Who we are reaching, and at what stage

What kind of ad each persona actually gets

What we claim × how we show iti

Read the rows before the cells

What we claim × how we prove iti

Where the evidence sits

What our ads actually doi

What moved since 2026-09-09

What has worked, and what has not

By form: what the ad does

By claim: what the ad says

The detail: each claim and form together

What already works

Where the biggest gaps are

We keep making the same recipe in batches, and Meta funds one ad per batch

Gifting is ready to rebuild for November: £55,146 at 2.01x last Q4

Value is the real empty row, and 12 months confirms it

Our one refill how-to video lost money when it opened the ad

Ritual ran four times, at the best return in the account

Press is empty on 30 days and on 12 months

What to make next

Prepare for Q4: last year's proven seasonal work

What needs deciding

What to build next

Method

What we claim × how we show it

What we claim × how we prove it

What our ads actually do