The decision, up front

Ailis wants three things: hold back on scaling until the autumn range lands (middle to end of next week, most likely Saturday 12 September), then push, and size the spend from her £257,511 September goal. The first two are right and the evidence backs them. The third does not survive contact with the ramp mechanics.

What her goal actually costs, and why it cannot be bought

Her £257,511 is not £257,511 of work for the ad account. £19,824 is already banked over days 1 to 3, and £92,803 arrives from the Recharge queue and returning customers whether we advertise or not. What is left, £144,884 of new-customer revenue, is the entire job, and on the measured response curve it prices at £104,922 of spend.

The account is on £1,060/day of Meta budget today. Getting from there to a £104,922 month means averaging £3,730/day across the 27 days left, and Meta will not absorb a jump like that: raise a budget more than roughly 20% and the ad sets re-enter learning, which is exactly what you do not want in a launch window. Here is the most the account can physically spend under clean 20% steps, holding the slow start she asked for:

Seven-day rungs are how we normally run, because each one produces a single clean CAC read. Four days is noisy but usable. Two days is not a read at all, and even that pace falls £8,648 short of her number while closing at £8,458/day.

The two Septembers we have on record

This is the part that should decide the size of the push, because we have run this experiment twice already and it came out the same way both times. Spend and revenue are Meta plus Google against Shopify orders, with new customers identified by first-ever order date rather than platform attribution.

September 2024 · autumn drop 18 September · August baseline £262/day at £34.38 CAC

September 2025 · autumn drop 23 September · August baseline £678/day at £31.42 CAC

One honest caveat on the 2024 read: it ran at £261/day against £993/day in 2025 and roughly £1,440/day now, so it is a signal about direction, not a level we can extrapolate. The 2025 year is the closer analogue and it is the one that says pushing hard into September is bought, not free.

She is right that September demand rises

Her premise was that last year was an anomaly because August was unusually strong. On the headline number that is not what happened: September has been below August three years running, so 2025 was not a one-off. But the composition says she is reading the market correctly even where the number disagrees.

The dip is existing customers, not acquisition. In both 2024 and 2025 new-customer revenue rose into September while returning and subscription revenue fell by a quarter. August pulls the list forward (August 2024 ran a 17.3% discount rate against 8.6% in September) and September pays it back. Nothing in that pattern says acquisition weakens.

UK Google Trends agrees, five years out of five. Index for "candles", monthly average:

Never once negative, +15% on average. And the seasonal term is sharper still: "autumn candles" peaks in September every single year and sits at zero from November through July. Its whole season is late August to the end of September, and it is gone by mid-October, which is an argument for going early and hard rather than late and cautious.

What we are launching into it

The autumn range lands 12 September on current expectation, 11 days earlier than last year's drop and six earlier than 2024's. All 30 autumn ads (832 to 861) are already live. The problem is not creative volume, it is that the creative has nowhere of its own to spend.

Across 1 to 3 September the autumn creative took £243 of £3,331, 7.3% of the account, because 8 of the ads sit inside the Evergreen CBO next to proven Tomato Vine winners where the algorithm starves them, and the dedicated Testing Autumn campaign is capped at £50/day.

So the ring-fence below is not a bet on autumn beating evergreen. It is buying enough conversions per week to get the autumn ad sets out of learning before the drop rather than during it, which is exactly what did not happen last year. Last September the dedicated Autumn campaign was not created until the 24th, the day after the drop, and it ran at £137/day while turning in the best ROAS in the account.

£400/day at the £30 CPA the autumn creative is actually running clears roughly 92 conversions a week, which is enough to exit learning in the eight days before the drop.

The three ladders

All three hold her slow start: £1,220 of Meta budget from the 4th and £1,400 from the 8th, a gentle +15% and +15% rather than a hold, then the real step on launch day. They differ only in how hard the back half of the month pushes.

The closing column is measured against £2,454/day, the highest off-peak day the account has ever run (9 August 2026). Every day above that figure in the account's whole history sits inside 15 October to 24 December 2025, when Christmas demand was carrying it.

Recommended: push to the £50 CAC stop line

This is the hardest push our own guardrail allows. It reaches £3,736/day, where forecast CAC hits exactly £50.00, and lands 86.7% of her goal at £21,044 of September profit. Set the Meta budget figure, never the total: Meta delivers about 94% of a configured budget and Google floats at roughly £310/day on top, which is a forecast rather than a decision.

The +81% move on the 12th is deliberate and it is the one step that breaks the 20% rule. Take it, because launch day resets learning anyway: new campaign structure, new creative, new destination. Every step after it is +20% or under, and each rung runs long enough to produce a usable CAC read before the next one.

How to run it

• Today, Fri 4 Sept. Set Meta to £1,220 and open the autumn ring-fence at £400/day in the dedicated campaign. Fix the autumn ad destinations before the budget steps up, or the 12th scales into a copy-to-page mismatch.
• Mon 8 Sept. Step to £1,400. Check the autumn ad sets have left learning.
• Drop day, 12 Sept. Step to £2,530. This is the one big move. If the drop slips, the step slips with it: every day of delay costs about £450 of spend capacity at the same closing budget, or forces a higher close.
• Gate on the 19th. This is the real decision point, not the 12th. Read trailing CAC on the 12 to 18 rung. Under £47, step on to £3,040 as planned. Over it, hold and finish the month on Plan C's close of £2,848/day, which is the peak already signed off.
• Hard stop. Seven days above £50 CAC means hold the rung, whatever the calendar says. The guardrail is CAC, never blended ROAS: blended slides down through any ramp by arithmetic, against £92,803 of base revenue the ads never touch.
• Watch the split, not just the total. If CAC holds but NC ROAS falls, that is basket size, and it is the autumn range's known weak spot. Fix the destination and the offer, do not cut the budget.

The month day by day

Recommended plan. Days 1 to 3 are actuals. Revenue is shaped by her email calendar and the drop, which the ads do not drive, so the days where spend and revenue deliberately diverge are visible. Her daily targets are her own revised sheet.

Method and sources

Revenue and customer splits are Shopify order data, 95,406 orders from November 2020 to 3 September 2026, with new customers identified by first-ever order date rather than platform attribution; recurring subscription orders are separated from one-off purchases. Spend is the Meta and Google Ads APIs at day level (Meta's window reaches back 37 months, so 2024 and 2025 are covered and 2023 is Google only). Search demand is UK Google Trends, weekly, aggregated to months; Google Ads Keyword Planner is not available on our developer token. The response curve is new-customer revenue per day = 6.212 × spend^0.8239, the same curve as the signed-off September plan, with £72.93 new-customer AOV, 56.5% contribution margin and £28,896 monthly overhead. As a cross-check, Shopify puts closed August 2026 at £199,589 against Triple Whale's £200,711.