Built 2026-07-29 for a 2026-07-30 launch, week CW31.
Five designs from the peppers autopsy, each getting the thing none of them ever had:
its own ad set, its own budget, and nobody else competing for it.
Every number below came from a read-only pull. Nothing has been created in Meta.
These are instructions for Tim, not actions taken.
0 / 0 steps done
How this pilot is governed
Read this once. It is the whole contract, and it is
what stops the pilot turning into another container full of creatives nobody can judge.
Delivery
Create every ad set and every ad PAUSED.
Tim switches them on by hand. Iris never activates an ad, ever.
$8.00 per day each, five ad sets,
$40.00 per day total. This is lane 1, the audition lane.
Stall test
Day four, 2026-08-02. An $8.00/day ad set
should be near $32 by then. Under $19 — less than 60% of its own budget
— it is stalled, and a stall is a decision owed. Pause it and re-slot the design
once, next Monday, as a brand new ad set with different creative. A second stall shelves it.
Audition gate
$56. At $8.00/day that arrives about
2026-08-05. The gate asks one question and only one: did it convert anyone at all?
Zero purchases at $54 is an automatic kill. Do not look at click cost. Do not look at
click-through rate. Do not think about the creative.
First re-verdict
Monday 2026-08-03.
Be precise about what this Monday can and cannot do: at roughly $32 spent, every ad set is
still TESTING, and a TESTING ad may not be decided. So 2026-08-03 is a
stall sweep and a delivery check — impressions above zero, nothing disapproved,
money actually moving. No return-on-ad-spend reading. No economic verdict.
Verdict Monday
2026-08-10. A Thursday launch puts the
$56 gate mid-week, so its verdict gets collected at the following Monday ritual.
Survivors go from $8.00 to $14.00 a day in place — raise the number and walk away.
No new ads, no duplicated ad sets, no creative edits.
Tracking
All five enter the daily digest.
The manifest at jobs/ads/pilots/oneadperadset-2026-07-30.json
lets tracking find them by ad set name before the real identifiers exist.
The moment Tim creates them, paste the real ad set IDs into that file — a name is
a fallback key, never a primary one, and this account already has two live ad sets sharing
one name.
Three things to settle before you click Create
1. The number 72 is already taken. The proposed names started at 72, but ad set
120248308217960726 is live right now carrying the name
72_Plan-V1-Mixed-Concepts-04_Flatlay_CW30. The cards below use
73 through 77 so the account's numbering stays unique.
2. Jimmy Nardello has already failed the zero-purchase kill. It spent $54.90
inside the peppers ad set alone with zero purchases, which crosses the $54 line. Putting it back
on the field is a deviation from the rulebook and needs a signature and an expiry date before it
happens, not after. The ready-to-sign wording is on card 4. If Tim will not sign it, build four
ad sets and promote the first reserve.
3. Reusing a post freezes where it sends people. Using an existing post is what carries
the likes, comments and shares across — but the click destination is baked into that post and
cannot be edited. Two of the five point somewhere other than the rulebook's answer.
Held as-is on purpose: this pilot changes one thing, the ad set shape. Changing where the
clicks land at the same time would leave you unable to say which change did what. The gaps are
written on each card and go in the queue as their own question.
One conflict between the written rule and the live account
The rulebook fixes every dedicated test ad set to open United States targeting.
Every live test ad set in the account targets 36 countries — the United States plus
Canada, Australia, New Zealand, the United Kingdom and most of the European Union. That is not a
small drift; it is a different audience.
The cards below follow the rulebook: United States only, and this is called out so it is
a choice rather than an accident. If Tim would rather match live practice exactly, that is
defensible — but then it is a deviation from the written rule and wants a dated line in
jobs/ads/decisions/ like anything else.
Read live on 2026-07-29 from ad set
71_Plan-V1-Mixed-Concepts-03_Flatlay_CW30 (120248289099250726),
the cleanest and most recent testing ad set in the account — created 2026-07-21, no
leftover custom audiences. Mirror it exactly, with the two marked exceptions.
Do not vary any of this between the five. Varying it turns a design test into a tangle of
confounded variables.
Campaign
20_ABO_TEST_10.07.2025 —
120232440887150726
Objective
Sales / conversions. Optimise for Purchase.
Never traffic, never a cost-per-result goal.
Pixel
425056660534207, conversion event
Purchase
Budget
Daily, $8.00 — this is the one field that differs from the
template, which runs $25.00
Bidding
Highest volume, no cost cap and no bid cap. Testing ad sets stay
uncapped — caps barely work without purchase history behind them, and this is the layer
with the least.
Attribution
7-day click, no view credit. Written down here before launch
and never changed mid-programme.
Countries
United States only — the second deliberate difference from
the live template, which runs 36 countries. See the flag above.
Age
18 to 65
Audience
Advantage audience ON. No interests. No custom audiences.
The template carries none, and the older peppers ad sets carried six — leave those behind.
Placements
Automatic / Advantage+ placements, as the template has them
Schedule
No end date. Standard pacing.
Ads per ad set
Exactly one. That is the entire point of the pilot —
the only ad set size at which Meta cannot starve anything is one.
SLOT 1Signal · proven family
Professional Tomato Plant Enabler
Why it earned the slot: it produced more purchases and more revenue than any
other ad in the peppers ad set — three orders and $193.59 on $114.30 — and it was still
spending money on the day somebody switched it off.
spend $114.30purchases 3revenue $193.59return 1.69add-to-carts 5per order $38.10
Use existing post — paste this
385785784617534_122197530602393295
Source ad 120248071826890726,
creative 3330849190429948, currently paused.
Reusing the post carries its existing likes, comments and shares into the new ad set.
Build a fresh ad and it starts at zero.
Destination — gap, held unchanged
Where it sends people todayinspiredgardenlife.com/collections/tomatoes
What the rulebook wants/products/professional-tomato-plant-enabler-t-shirt-3214173
It advertises one specific design, and a specific design belongs on its exact
product page — not a collection. The Tomatoes collection is a registered, approved
path, so this is a gap and not a violation. Do not change it this round.
Colour: nothing to preselect — the best-selling colour is Black at only 25% of twelve units
over 120 days, which is far too thin to force.
SLOT 2Signal · proven family
Quietly Cheering for Moss — Military Green
Why it earned the slot: it posted the highest return of any creative in the
autopsy that sold more than once — 1.85 against a 1.87 break-even, which is
effectively paying for itself — and then somebody switched it off after two days.
Source ad 120248071826830726,
creative 1672959090425752.
It also ran as 120247996869850726 in the other peppers ad set —
same creative, same post, so the engagement from both runs is already pooled on this one post.
Destination — the widest gap of the five, held unchanged
Where it sends people todayinspiredgardenlife.com/collections/seeds
What the rulebook wants/products/quietly-cheering-for-moss-t-shirt-4182582?variant=47749536973015
A moss shirt is currently pointing at the Seeds collection. The path is
registered and approved, so nothing is broken — but the subject does not match, and a
shopper who clicked because of moss arrives at seeds. This is the strongest candidate for the
next destination experiment.
Colour is the happy part: the ad shows Military Green, and Military Green is this product's
number one seller by a mile — 186 of 258 units over 120 days, 72%. The variant link
above opens it on Military Green. Still, do not change it this round — it earned 1.85
with the destination it has, and the pilot is testing shape, not landing pages.
SLOT 3Signal · proven family
Season of Patience — Limited Advent Collection
Why it earned the slot: the best all-round record in the whole autopsy
— it bought twelve add-to-carts at $6.89 each, the cheapest of any ad with real money
behind it, and turned two of them into orders at a 1.67 return.
Source ad 120248071826850726,
creative 2155417555025063, headline
“Wear the Season's Quiet Wisdom”.
This one also has a live Instagram post attached —
instagram.com/p/DbGnQrtAjsi/ — so there is real social proof on
both sides to carry forward.
Destination — already correct, no change
Where it sends people todayinspiredgardenlife.com/collections/simple-living
What the rulebook wants/collections/simple-living — the same thing
This one advertises a collection, not a single design, so there is no single
product page to send it to and no advertised colour. A seasonal or theme campaign belongs on its
matching registered collection, and Simple Living is a registered primary collection.
Nothing to fix.
The obvious objection, answered
Yes, it is an Advent ad and it is July. Worth saying out loud so
nobody has to wonder. But it produced that record in July — it spent, carted and
converted between the 21st and the 23rd of this month, out of season, and still came closest to
break-even of anything in the autopsy. It also carries no expired discount in its copy,
which is exactly what disqualified the Valentine's grid ad. The season is a reason to watch it,
not a reason to bench it.
SLOT 4New · never fairly tested
Jimmy Nardello
Why it earned the slot: it is the one real pepper design that got any
traction inside an ad set that was named for heirloom peppers and then spent its money on
Valentine's mock-ups and a tomato joke. Nine add-to-carts at $8.06 and 233 clicks. People
want the shirt and then stop — that is a specific, findable failure, not a dead design.
Jimmy Nardello has already failed the zero-purchase kill. It spent $54.90 inside
ad set 120248071826590726 on its own, with zero purchases, and the rule
says $54 with no sales is an automatic kill. Re-slotting it is a deliberate deviation, and a
deviation only counts as an override if it is written down before it happens, with an
expiry date. Otherwise it is just the rulebook quietly changing.
Copy this into
jobs/ads/decisions/2026-07-30-jimmy-nardello-k1-override.md and sign it:
What changes: hard kill K1 is suspended once for the
design Jimmy Nardello, which has already crossed $54 with zero purchases. Why: it never once ran as a single-ad, single-budget test — it shared a $25/day ad set
with 33 other creatives, and the evidence against it was bought under a structure this pilot exists
to replace. Who agreed: Tim — ____________________ Expires: 2026-08-10. The number that judges it, chosen now and not later: at least one purchase by $56 of
spend in the new ad set. Zero purchases at $56 is a kill under K1, and there is no second
reprieve.
If Tim will not sign this, do not build card 4. Build four ad sets and
promote Quietly Plotting from the reserves instead.
Use existing post — paste this
385785784617534_122197514738393295
Source ad 120248071826740726,
creative 2249689715800597, headline
“This Pepper Has Ancestors”. It also ran as
120247577210410726 — same creative, same post, engagement pooled.
Instagram post: instagram.com/p/DalozsmAKAJ/
Destination — already correct, no change
Where it sends people todayinspiredgardenlife.com/products/jimmy-nardello-shirt
+ campaign, ad set and ad name tracking tags
What the rulebook wants/products/jimmy-nardello-shirt — the same thing
One of only two of the five that already lands exactly where it should
— the precise product page for the design being advertised, with tracking attached.
No colour to preselect: the ad name carries no colour and the product has no colour history worth
acting on. Which makes the $8.06 add-to-cart and zero purchases genuinely interesting
— the destination is right, so whatever is leaking is downstream of the product page.
SLOT 5New · never fairly tested
More Wildflowers, Less Hurry
Why it earned the slot: it posted the highest return in either peppers ad set
— 3.08 — on $12.29 of spend. That is one order, so statistically it proves
nothing at all. It is here precisely because it has never had a real budget in its life, and
it is the most promising untested thing in the autopsy.
Source ad 120248071826730726,
creative 1006537815343797, headline
“More Wildflowers. Less Hurry.” It also ran as
120247577233540726 — same creative, same post.
There is not much engagement on it yet, which is the whole point: it has barely been shown to
anybody.
Destination — already correct, no change
Where it sends people todayinspiredgardenlife.com/products/more-wildflowers-less-hurry-shirt
+ tracking tags
What the rulebook wants/products/more-wildflowers-less-hurry-shirt — the same thing
The second of the two that already lands exactly right. No colour to preselect
— the ad name carries none, and the product's colour history is six units, which is nothing.
So this slot is as clean a read as the pilot can produce: right page, right product,
one ad, its own money, nothing else to blame.
BEFORE YOU WALK AWAY
The two that earned a slot and did not get one
Both were on the autopsy's list. Neither is being dropped
— here is exactly why each sat this round out, so nobody has to re-argue it later.
First reserve — Quietly Plotting. $96.08 spent, 2 purchases, eleven add-to-carts
at $8.73 and 256 clicks. Its front-of-funnel plainly works. It is clean under every hard kill.
Why it sat out: it is the only one of the seven whose creative has
no website destination at all — it sends people to the shop inside Meta rather than to
the storefront. Slotting it would mean changing the ad set shape and the destination in the same
week, and then nobody could say which change moved the number. Promote it immediately if the
Jimmy Nardello override is not signed.
Post ID 385785784617534_122128966196393295 ·
product quietly-plotting-my-next-tomato-harvest-t-shirt-1975430
Blocked, not deprioritised — the Valentine's 2x2 grid. $50.55 spent, no purchases,
nine add-to-carts at $5.62 — the cheapest of anything with meaningful spend, and both
copies only started delivering on 2026-07-28.
Why it cannot be used: the live post copy reads
“30% off all shirts through Valentine's Day” and its description asset reads
“Valentine's Sale — 30% off.” Reusing that post in July advertises a
discount that does not exist. That is a customer-trust problem, not a seasonality preference.
It cannot be re-slotted as an existing post at all — the copy has to be rewritten
first, which means a new post and no inherited engagement. Queue it as new creative for a
seasonally sensible window.
Post IDs 385785784617534_122197516058393295 and
385785784617534_122197547954393295