Meta treats near-identical ads as one ad. Ten lookalikes get one shot at reaching people; ten different ideas get ten. This page checks each live ad set, then the account, then says what to make next.
01
Ad sets
Every ad set live today with spend between 24 Sept 2026 to 23 Oct 2026. Our rating uses the rule in Method. Meta's own "Creative diversity" column (Low / Medium / High, per ad set, in Ads Manager under Columns, Customise columns) is not yet readable by API, so that column is blank here: add it from Ads Manager when you look.
Ad set
Our rating
Meta's rating
Why
Do
Christmas - Broad - Exclude Customers
Cold & Warm | Evergreen (old Tomato vine)
£26,347 in the window · 15 of 38 ads getting real delivery · 4 distinct ideas
Low
–
96% of the money makes one claim: gifting / occasion.
Add the next-best argument from the account's own history next to it (see Build next).
Everything Else - Broad - Exclude Customers
Cold & Warm | Evergreen (old Tomato vine)
£16,350 in the window · 12 of 13 ads getting real delivery · 6 distinct ideas
Low
–
92% of the money runs on video.
Add the missing format on the same winning argument: a static version of what already works here.
Tomato - Broad - Exclude Customers
Cold & Warm | Evergreen (old Tomato vine)
£13,079 in the window · 16 of 26 ads getting real delivery · 6 distinct ideas
Medium
–
Every ad here speaks to one buyer: the scent-led home curator.
Add creative for the next persona in the persona doc, in this same ad set. Never a separate ad set for them.
Seasonal - Broad - Exclude Customers
Cold & Warm | Evergreen (old Tomato vine)
£11,090 in the window · 19 of 24 ads getting real delivery · 4 distinct ideas
Medium
–
62% of the money makes one claim: sensory / product desire. Working, but narrow.
Add the next-best argument next to the winner (see Build next).
Ad Set 1 - MAIN - Broad - Exclude Customers
Cold & Warm | Cook Candles
£7,566 in the window · 15 of 98 ads getting real delivery · 6 distinct ideas
Medium
–
Every ad here speaks to one buyer: the scent-led home curator.
Add creative for the next persona in the persona doc, in this same ad set. Never a separate ad set for them.
Gifting - Broad - Exclude Customers
Cold & Warm | Gifting Christmas 2026
£6,385 in the window · 8 of 14 ads getting real delivery · 3 distinct ideas
Low
–
78% of the money makes one claim: gifting / occasion.
Add the next-best argument from the account's own history next to it (see Build next).
Ad Set 1 - Customers L180D + LCA365D Excl Recent/Subs | Opt SubscriptionPurchase
£2,069 in the window · 9 of 17 ads getting real delivery · 3 distinct ideas
Why. 68% of the money makes one claim: subscription / value ladder. Working, but narrow.
Do. Add the next-best argument next to the winner (see Build next).
Fir A/B - Statics - Broad - Exclude Customers
Cold | New Customers | Fir A/B Test Statics Christmas 2026
Too young
£1,350 in the window · 4 of 4 ads getting real delivery · 2 distinct ideas
Why. Live for 6 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 31 Oct 2026.
Fir A/B - Videos - Broad - Exclude Customers
Cold | New Customers | Fir A/B Test Videos Christmas 2026
Too young
£917 in the window · 3 of 3 ads getting real delivery · 3 distinct ideas
Why. Live for 6 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 31 Oct 2026.
Test 3 Starter Pack Scent - Broad - Exclude Customers
Cold | New Customers | Christmas Creative Testing 2026
Too young
£679 in the window · 1 of 4 ads getting real delivery · 1 distinct idea
Why. Live for 2 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 4 Nov 2026.
Test 4 Icicle Proof - Broad - Exclude Customers
Cold | New Customers | Christmas Creative Testing 2026
Too young
£653 in the window · 2 of 4 ads getting real delivery · 1 distinct idea
Why. Live for 2 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 4 Nov 2026.
Test 1 Fir Proof - Broad - Exclude Customers
Cold | New Customers | Christmas Creative Testing 2026
Too young
£616 in the window · 2 of 4 ads getting real delivery · 1 distinct idea
Why. Live for 2 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 4 Nov 2026.
Test 2 Fir Message - Broad - Exclude Customers
Cold | New Customers | Christmas Creative Testing 2026
Too young
£466 in the window · 3 of 3 ads getting real delivery · 1 distinct idea
Why. Live for 2 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 4 Nov 2026.
Klaviyo All Emails - Zero Orders L90D | Excl Subscribers
Hot | Retention | All Purchasers | Christmas 2026
Too young
£215 in the window · 1 of 16 ads getting real delivery · 1 distinct idea
Why. Live for 4 days. Too early to judge; Meta needs a couple of weeks of delivery before the pattern means anything.
Do. Rate from 2 Nov 2026.
Two rules that do not moveDiversity is fixed by making different creative, never by splitting delivery into stage or persona ad sets. And "saturated" means stop making more of a lane, never stop spending on it; Meta decides where the money goes.
02
Three questions about the account
Spend-weighted, every ad that spent in the window (288 ads, £87,901).
Are we talking to every buyer type?
partly
Scent-led home curator 52%, Refill loyalist 3%, Gift-giver 38% of spend, 6% not readable from the ad. The refill loyalist is under 10%.
Are we meeting them at every stage?
partly
cold (needs teaching) 43%, comparing 45%, ready to buy 5%, 6% not readable. Ready to buy under 10%.
Do the ads look and sound different from each other?
partly
Formats video 48%, static 51%, carousel 1%. 15 kinds of content live, top 33%. Who carries the message: nobody on screen (product only) 76%, a customer 21%, founder 2%, a partner page 0%. Three in four pounds run with nobody on screen.
03
Build next
Five cells, each an argument crossed with a content type, taken from the 20 Oct ranked list and capped to what can be shot in the next two weeks. Each empty lane gets two to four executions, so a miss tells us about the lane and not about one ad. None of these needs its own ad set; they go into the single testing campaign.
1 Value / economics × Text-led static + carousel
What a year of candles costs
Why the lane is empty. No ad in 365 days has put a pound figure on the saving. Ad 922 implies value but never states a price and has the lowest conversion rate of the big spenders (2.13% on £10,333). Comparison devices are 0.1% of spend.
Why it could work. The one ad that did state a number, Ad 255 ("refill worth £27.50"), spent £4,196 at 2.59x with a £29.8 cost per sale. Vague value claims lost: Ad 367 "your candle habit costs less" ran 0.52x to 1.19x.
Make. A text-led static ("Pott, once. Candle refills, four a year. Per evening: ..."), a carousel with one refill per card and the running total on the last, and a two-person kitchen-table video doing the maths. Seasalt or Tonka Standard Starter Pack, not Fir.
2 How it works, proved by a customer × Review card on an informal photo
The refill, in a customer's words
Why the lane is empty. A third of spend states a claim with no proof, and the how-it-works claim is almost entirely asserted by the brand rather than shown by a customer.
Why it could work. Ad 1029, a review on an informal phone photo, ran 3.16x with an £18.0 cost per sale on 8 purchases. The Claire quote ads 205 and 217 did 3.13x and 1.89x on £5,191.
Make. Three real Yotpo reviews about the refill or keeping the Pott, each on a phone photo taken on the day. Gingham or house Potts with Starlight, Fig or Seasalt refills.
3 Sensory / the scent × Scent-story static + hands and voiceover cut-down
Scent notes, in order
Why the lane is empty. The scent claim fell from 18.9% to 1.9% of spend in a month while the account moved to gifting. It is the biggest review theme (45% of reviews) and the core self-buyer's live risk ("cannot smell it online").
Why it could work. Scent-led teaching worked at 2.29x on £8,946 before it was switched off 118 days ago; Ad 554 Roses did 2.28x on £4,316.
Make. One static scent story per scent (top note first, then the base, in Ailis's plain voice) plus a 15-second hands-and-voiceover cut. Starlight, Fig and Noël, not Fir.
4 Objection handling × Founder answers real comments
Ailis answers real comments
Why the lane is empty. Objection handling has had no ad in 365 days. Price is the top comment objection and the founder's share of spend fell from 11.1% to 2.2%.
Why it could work. Founder-captioned Ad 483 did 2.85x on £3,118 with a £23.1 cost per sale. Caution: the Fir founder video 1031 is at 0.88x on £115, so keep this off Fir until that reads.
Make. One long piece and three cut-downs: founder's hands and voice over a real comment sticker, answering straight away with no lead-in. Needs three or four real comments screenshotted first.
5 Ritual × Sound-led macro, no voice
Just the sounds
Why the lane is empty. The ritual claim has had nothing live in 30 days and only 0.2% over the year, yet it sells the system rather than one scent.
Why it could work. Ritual ads returned 2.77x on £1,624 over the year; the match-strike Ad 132 did 2.39x on £1,192 and 3.96x on £416.
Make. Three 15-second close-ups with no voice: the match on the MatchPott, the wick catching, a refill dropping into the Pott. One phone, a quiet room, about an hour. Noël or Starlight refills.
Not on the list, on purpose: gifting review cards on Holly and gifting showcase statics. They are saturated (22 working ads carry 32% of spend). Keep spending on them; stop adding more of them.
Method
What counts as a different idea
Two ads are different concepts only if they differ on at least two of: who it is for (the persona), what it claims (the argument), and what it looks like (the content type). A new creator reading the same script, a headline swap or a different scent in the same shot is the same concept. Iteration is healthy (roughly 70% of a good account is iteration); the failure is an account that is all iteration and calls it testing.
The rating rule, per ad set
Count the ads getting real delivery (£100+ in the window) and the distinct concepts among them.
High: 4+ ads working, 3+ distinct concepts, no single format over 85% of spend, no single claim over 60%.
Low: under 4 ads working, or one concept, or any of the triggers below.
Medium: everything else.
Too young: live under 10 days. Rated two weeks after launch.
Triggers, first match wins, each with its fix. Low (Meta has little to choose between): too few ads working, under 4 (add executions); one concept only (build a second and third); all one look, the top three ads share a content type and messenger (change the messenger or setting); all one format, over 85% of spend (add the missing format on the winning argument); one claim over 75% of spend (add the next-best argument). Medium (working, but narrow): one claim at 60 to 75% of spend; every ad to one persona (add creative for the next persona, same ad set); all one stage, nothing for cold viewers or nothing for ready buyers (add that stage as creative); only two distinct ideas. High: none of the above. "All one look" is judged from the content-type tags, not from the thumbnails themselves.
Why the checklist is fixed
An ad that was never made leaves no trace in the data. Every ad is mapped onto a fixed list of 14 arguments and 23 content types, so an empty lane is visible. Where the ad's name does not say what it argues, the first 400 characters of copy are read; for the argument, 148 ads were tagged from their labels, 69 from copy and 66 fell back to the catch-all.
Arguments, last 30 days
Claim
Share
Working ads
Us vs them / anti-category
3%
4
Gifting / occasion
32%
22
Press / PR / authority
0%
0
Social proof / testimonial
5%
9
Founder / craft / origin story
13%
14
Value / economics
18%
11
Features & benefits
1%
3
Problem, solution
0%
1
Education / how it works
1%
3
Objection handling
0%
0
Subscription / value ladder
4%
8
Identity / ritual / aspiration
0%
1
Sensory / product desire
23%
32
Content types, last 30 days
Type
Share
Working ads
Whitelisted / partnership ad
0%
1
Founder video / talking head
2%
6
Unboxing
10%
10
UGC testimonial
5%
8
UGC demo / first use
6%
12
Review / press screenshot graphic
20%
12
How-to / tutorial
33%
24
Behind the scenes / making-of
3%
1
Lifestyle video / silent b-roll
5%
8
Lifestyle static
4%
6
Product / studio static
9%
13
Social-native / story style
0%
2
Problem-solution video
0%
0
Meta's rating
Meta added a "Creative diversity" column to Ads Manager in late August 2026. It rates each ad set Low, Medium or High on how visually different its images and videos are, and is still labelled "in development". Meta has not published the method, and agencies report mixed-format ad sets still scoring Low. Treat it as a flag that something looks the same, not a verdict; this page says what the "something" is. The field is accepted by the reporting API but returns no value yet, so it has to be read in Ads Manager.
Data
Ad-level spend 24 Sept 2026 to 23 Oct 2026 from the Meta reporting API; creatives and names from the 20 Oct 2026 audit pull (777 ads); account lanes from the audit analysis of the same date. Persona and stage are inferred from the argument and messenger using the persona doc's rules, so they are indicative. Nothing on this page changes anything in the ad account.