Pott Sep 2026 Forecast Tracker

What this is

Two September forecasts scored daily against real results, both frozen before the month started, so the whole month is a fair race. The model is the sep-2026-plan ramp (£69,650 spend landing near £215,657 = 83% of her goal), day-shaped by the Recharge billing schedule for subscriptions and by measured weekday and email-day effects for everything else. The client sheet (£260,511) is scored exactly as she built it. Her email calendar drives both lines' day shape, so a launch day that lands flat hits both forecasts, not just hers.

Month landing view

	Revenue	Ad spend
Model landing (actuals + remaining model forecast)	£215,658	£69,650
Client sheet landing (actuals + her remaining plan)	£260,511	n/a (no ad plan in her sheet)
Client sheet full-month goal	£260,511	n/a
Model full-month forecast	£215,658	£69,650
Actual to date (complete days + today so far)	£0	£0

Guardrail: the ramp steps up only while 7-day CAC holds under £50. CAC could not be refreshed on this run, so no number is quoted rather than a stale one.

Forecast accuracy scoreboard

MAE = average daily error. Bias = cumulative forecast vs cumulative actual (positive = forecast too high). The online rows are the interesting ones: they test whether the measured email-day lifts (regular email +£553, launch +£4,118, weekday-controlled, 152 days of evidence) predict her calendar's spikes better than her round-number day targets.

Forecast	Days scored	MAE / day	Cum bias	Cum forecast	Cum actual

Why the two forecasts disagree (£260,511 vs £215,658)

Same VAT basis, same revenue definition - the gap is real, and it splits into two parts.

Component	Her sheet	Model	Gap	Why
Subscriptions	£63,761	£55,931	−£7,830	Her daily subs line is the same 30 numbers pasted by day-position into every month Sept to Dec (closest to June's shape). The model line is the actual Recharge billing schedule: 97% of September is already queued with a date and an amount.
Online (new + returning)	£196,750	£159,727	−£37,024	Her online line is a top-down day-spread of the goal (£6,558/day average). The measured no-email baseline is ~£3,100/day of non-sub revenue; closing her gap is 80% more new-customer revenue than August produced, and her sheet contains no ad plan to buy it.

Her email-day jumps are directionally right and are kept in the model's day shape: regular email days really do run about +£550 and launch days about +£4,100 over the weekday baseline (best observed: Emily Doran shop-now day, +£10,200). Her £14,000 target for the 12 Sept Autumn Launch needs a best-ever collab-scale day; the model carries that day at £13,352 - still its biggest day of the month.

Daily detail

Spend plan is the approved ramp: £1,750 → £2,100 → £2,500 → £2,800 total/day in weekly steps (Meta budgets to SET are grossed up by the 0.94 delivery ratio: £1,606 / £1,979 / £2,404 / £2,723). Actual spend is Meta plus Google. Cum Δ columns: actual minus forecast, running. Greyed rows are future days; today is excluded from scoring until complete.

Ad spend Subscription revenue Online revenue Total revenue Cumulative variance Bright text = actual · dim text = forecast

Day	Calendar	Ad spend		Subscriptions			Online			Total revenue			Cum Δ
Day	Calendar	Plan	Actual	Model	Client	Actual	Model	Client	Actual	Model	Client	Actual	Model	Client
1 Tue	Scent of the month: Fig (launch email)	£1,750		£3,948	£4,848		£5,919	£7,000		£9,867	£11,848
2 Wed		£1,750		£1,283	£1,769		£4,102	£5,000		£5,385	£6,769
3 Thu	Cook / back-to-school email	£1,750		£1,558	£1,818		£4,582	£7,000		£6,140	£8,818
4 Fri		£1,750		£1,635	£2,058		£4,016	£5,000		£5,651	£7,058
5 Sat	Fig Reviews email	£1,750		£1,533	£1,808		£4,759	£7,000		£6,292	£8,808
6 Sun		£1,750		£1,696	£2,065		£4,416	£5,000		£6,112	£7,065
7 Mon	Last Chance Citronella & Roses email	£1,750		£1,642	£1,882		£4,626	£5,000		£6,268	£6,882
8 Tue		£2,100		£1,011	£1,273		£4,529	£7,000		£5,540	£8,273
9 Wed	Autumn is coming email	£2,100		£1,513	£1,837		£5,033	£5,000		£6,546	£6,837
10 Thu		£2,100		£1,714	£1,923		£4,512	£5,000		£6,226	£6,923
11 Fri		£2,100		£1,746	£1,819		£4,446	£4,500		£6,192	£6,319
12 Sat	AUTUMN LAUNCH: Rust & Amber (her sheet: £14,000 day)	£2,100		£1,382	£1,497		£11,970	£14,000		£13,352	£15,497
13 Sun	day after launch	£2,100		£1,575	£1,857		£6,212	£7,000		£7,787	£8,857
14 Mon		£2,100		£2,156	£2,318		£4,556	£6,000		£6,712	£8,318
15 Tue	Meet Amber email	£2,500		£2,502	£2,782		£5,506	£6,250		£8,008	£9,032
16 Wed		£2,500		£1,070	£1,259		£5,009	£6,500		£6,079	£7,759
17 Thu	Ways to Style your Potts email	£2,500		£1,382	£1,544		£5,489	£6,500		£6,871	£8,044
18 Fri		£2,500		£1,794	£1,797		£4,923	£6,000		£6,717	£7,797
19 Sat	Three perfect autumn scents email	£2,500		£1,714	£1,920		£5,666	£6,000		£7,380	£7,920
20 Sun		£2,500		£1,641	£1,988		£5,323	£8,000		£6,964	£9,988
21 Mon		£2,500		£1,314	£1,441		£5,032	£7,000		£6,346	£8,441
22 Tue	Saying goodbye to Diffusers email	£2,800		£1,372	£1,665		£5,855	£8,000		£7,227	£9,665
23 Wed		£2,800		£1,585	£1,790		£5,358	£6,000		£6,943	£7,790
24 Thu		£2,800		£1,385	£598		£5,337	£8,000		£6,722	£8,598
25 Fri		£2,800		£1,827	£2,103		£5,272	£5,000		£7,099	£7,103
26 Sat		£2,800		£1,277	£2,067		£5,514	£8,000		£6,791	£10,067
27 Sun		£2,800		£1,757	£2,588		£5,672	£6,000		£7,429	£8,588
28 Mon		£2,800		£3,620	£3,684		£5,381	£6,000		£9,001	£9,684
29 Tue		£2,800		£2,159	£2,191		£5,354	£8,000		£7,513	£10,191
30 Wed		£2,800		£5,140	£5,572		£5,358	£6,000		£10,498	£11,572

How the model's day shape is built

Subscriptions: the per-day Recharge queue for September (real scheduled charges), scaled to the plan's ledger expectation of £55,931 (queued × 0.93 realization plus the small roll-forward remainder).
New-customer revenue: flat within each ramp rung (spend × that rung's expected NC ROAS from the measured response curve, b = 0.822).
Returning non-sub (£49,500): weekday-shaped on the no-email baseline measured Apr-Aug (Sun highest at ~£3,670, Fri lowest at ~£2,836).
Email-day lifts, measured weekday-controlled on 152 days Apr-Aug 2026: regular email +£553 (54 days), scent-of-month launch ~+£2,000, collab-scale launch +£8,000 (12 Sept), day-after halo +£1,500. The whole online curve is then rescaled so the month sums to the ramp landing - lifts shape days, the response curve sets the total.

Method notes

Actuals: Shopify order total_price minus refunds, bucketed by created-at day (Europe/London). Subscriptions = orders tagged "Subscription Recurring Order" or sourced from contract billing. Spend = Meta account daily + Google Ads v25 daily.
Both forecast lines frozen 31 Aug 2026 from build-forecast-data.cjs; the nightly run only adds actuals and re-scores.
Client sheet source: "POTT forecast Aug - Dec 26" workbook ("May targets"), September tab, incl VAT - same basis as all actuals here.
Watch-outs carried from the plan: the 12 Sept Autumn Launch carries 7% of her month's online target on one day; Summer Sale pull-forward may depress days 1-2; scoring the ramp against constant-ROAS expectations will make it look like it is underperforming (use this page's model line, not the £223,618 old number).
Rerun: node update-tracker.cjs in this folder.