capacity-and-covers-forecast

---
name: capacity-and-covers-forecast
description: Builds a covers and staffing forecast from the venue's own booking history, checked against the current direction of the national eating-out market, rather than from last year's rota copied forward. Reads whether the sector is growing or shrinking from the latest published trackers, and flags where the venue's own trend moves against the published direction, the point where a forecast built on habit stops being safe to trust. Use when setting covers, staffing or stock for an upcoming period, or checking your trend against the wider market.
---

# Staff and stock for the covers actually coming

Give it the venue's covers and sales for recent weeks. Get back a forecast for the coming period, checked against the latest published sector direction, with a flag wherever the venue's trend moves opposite to it.

## What it does

1. **Start from the venue's own recent numbers**, never a rota copied forward unchanged.

2. **Check that figure against the latest published sector tracker.** The NIQ RSM Hospitality Business Tracker reported restaurant groups' sales "down by 0.7% year-on-year in June", called "their second-worst month" of 2026.

3. **Use a second, independent tracker to confirm direction**, since one report can be a blip. Lumina Intelligence's panel found, for the four weeks to early February 2026, "eating out participation edged up to 57.9% (+0.4ppts year-on-year)", "visit frequency rose sharply (+7.3%)", and "average spend jumped +9.4% to £17.86 per visit."

4. **Name any gap, don't smooth over it.** Where the venue rises while the trackers show a falling period, or vice versa, record it as a flagged gap rather than quietly averaging the two.

5. **Turn the forecast into covers per shift**, with staffing and stock as a separate, explicit step built from that figure.

6. **Reforecast on a stated cycle**, not once a quarter from memory; both trackers used here publish monthly.

## Then it checks

1. The base figure is drawn from the venue's own recent history, with the date range stated.
2. At least one sector tracker figure is recorded alongside it, with source and date.
3. Any gap between venue and tracker direction is named explicitly.
4. The forecast is expressed as covers per shift, with staffing as a separate step.
5. A reforecast date is set.

Any check fails: redo that step once. Failed twice: say what is wrong and stop.

## Rules
- Public information only. Never invent a fact, number or quote. Anything sent in someone's name says whose name it is.
- Never present one tracker's figure as "the market"; use at least two independent readings.
- Never quietly blend the venue's trend into the sector figure; a gap is recorded, not averaged away.
- Working document for the owner to check against their own records. Not a guaranteed prediction.

## Built from
- NIQ RSM Hospitality Business Tracker, rsmuk.com/insights/hospitality-business-tracker, June 2026 trading, read 4 Nov 2026: step 2.
- FRY Magazine, reporting Lumina Intelligence's Eating and Drinking Out Panel, frymagazine.com, published 28 Apr 2026, read 4 Nov 2026: step 3.

# capacity-and-covers-forecast

## You are given
The venue's own covers and sales history for recent weeks, by date. The forecast period to plan for (dates). The latest published sector tracker figures, already supplied by Claude: tracker name, the year-on-year direction and figure, the date range it covers, and the date it was read. A second, independent tracker's figures in the same form, where available. The reforecast cycle (for example, monthly), set by the owner.

## Produce
Write into a `./capacity-and-covers-forecast-output/` folder:

1. `venue-trend.csv` with columns: `week_start`, `covers`, `sales_gbp`, `week_on_week_covers_change_percent`, `trend_direction`. `trend_direction` is `rising`, `falling` or `flat`, calculated from the covers trend across the supplied weeks.
2. `sector-trackers.csv` with columns: `tracker_name`, `producer`, `period_covered`, `direction`, `figure_verbatim`, `date_read`. One row per tracker supplied.
3. `forecast.csv` with columns: `forecast_period_start`, `forecast_period_end`, `base_covers_per_shift`, `basis_description`, `reforecast_date`. `basis_description` states in one line which weeks of venue history the base figure was drawn from.
4. `gap-check.csv` with columns: `venue_trend_direction`, `tracker_1_direction`, `tracker_2_direction`, `gap_found`, `gap_description`. `gap_found` is `yes` where the venue's direction differs from either tracker's direction, otherwise `no`.
5. `gaps.md` - a numbered list of: any `gap_found` row of `yes`, restated with the directions involved; any tracker row missing a `date_read`; any forecast with no `reforecast_date` set; any venue-trend row with a missing `sales_gbp` or `covers` value.

## Rules
- Codex calculates and records. It never decides the forecast's base covers figure beyond the stated calculation from supplied history, and never explains away a gap between the venue and the trackers - that explanation, if any, is supplied by the owner and recorded separately, not invented here.
- Never report a sector tracker figure without its producer, period covered and date read.
- Trend directions and gaps are calculated exactly from the supplied figures, never estimated.
- Use British English, GBP and DD Month YYYY dates. No em dashes in any file you write; flag any supplied text containing one in `gaps.md`.
- Every file ends with this line: this is a working document prepared for the owner to check against their own booking and sales records, and is not a prediction guaranteed to hold for any specific week.

## Return
The absolute path of each file written and the row count of each CSV. The venue's own trend direction and the figure it was calculated from. Each sector tracker's direction, figure and date read. The forecast's base covers per shift and its reforecast date. The gap-check result and description if a gap was found. The `gaps.md` item count.