Forecast-based prep

By default, prep list batch recommendations are computed from burn rate — historical consumption rate over a lookback window. Burn rate is a great signal for most kitchens, but it has a blind spot: it can't see ahead. If next Friday is a holiday, has unusually warm weather, or follows a marketing push, last month's average won't catch the spike.

Forecast-based prep flips the model: instead of "what did we use last month?", it asks "what will we sell tomorrow, and what ingredients does that require?"

How it works

When forecast mode is enabled on a prep list, Rinvy:

  1. Applies the stored demand forecast day by day over the prep window's actual dates (default: through the next prep date) — dates inside the forecast window use that exact day's predicted portions, dates beyond it use the item's per-operating-day average, and closed days contribute nothing.
  2. Multiplies those expected sales × the menu item's recipe ingredients × the portion multiplier.
  3. Rolls the ingredient demand up through any sub-recipes (recipe explosion).
  4. Subtracts current stock and divides by recipe yield to get recommended batches.

This is the same mechanism the order list's forecast mode uses: untouched items keep the forecast's exact per-day portions, and editing an item's expected sales rescales its day-to-day shape proportionally rather than flattening it (setting 0 removes the item).

The output is the same shape as burn-rate-based prep — batch counts per recipe — but the math is forward-looking.

Prerequisites

Forecast-based prep needs:

  • Pro tier for both prepLists and demandForecast feature flags.
  • Sales data uploaded — forecasts use sales history as the primary input. Without sales, the forecast falls back to burn-rate-derived estimates.
  • Menu items linked to recipes (forecasts are per-menu-item; recipes are how they get converted to ingredients).
  • A current forecast — see the demand forecast page to refresh.

Enabling forecast mode on a prep list

  1. Open the prep list while in DRAFT.
  2. Switch the recommendation source from Burn rate to Sales forecast.
  3. Review the new recommendations. They'll typically differ from burn-rate numbers — sometimes meaningfully.

  4. Adjust as needed and approve.

The mode is set per prep list, not per recipe. You can switch back to burn rate any time before approving.

When to use which

SituationUse
Stable weekly cycle, no upcoming changeBurn rate
Holiday weekend or known eventForecast
Recent menu change, no sales data yetBurn rate (forecast won't have enough history)
Weather-sensitive menu (patio, cold drinks)Forecast
First few weeks on RinvyBurn rate (forecasts need history to be useful)

Why the numbers can move a lot

Burn rate and forecasts can disagree significantly. A few reasons:

  • Burn rate averages across the lookback window; forecasts are date-specific.
  • Forecasts can adjust for weather, day of week, operator-entered context, and day notes logged during the history period.
  • Forecasts learn from their own track record — persistent over- or under-predictions (including per-weekday patterns) get corrected in the next generation.
  • With a weather location set, forecasts also weigh observed weather history — how your kitchen actually traded on rainy vs. dry and warm vs. cold days — not just the upcoming forecast.
  • A forecast for tomorrow says "we'll sell 40 burgers" — burn rate says "we sold 24 burgers/day on average over the last 30 days."

When they diverge, trust the source with more information about that specific day. For a normal Wednesday, burn rate is fine. On Mother's Day, the forecast captures the spike that the burn-rate average smooths away.

Common mistakes

Switching to forecast mode with an outdated forecast

The Outdated chip on the demand forecast page lists exactly why a forecast is behind — generated more than a day ago, changed business context, new sales data, changed menu items, or a window that has already passed. Regenerate before relying on forecast-based prep.

Expecting forecasts to be exact

A forecast is a prediction with uncertainty. Use it as a starting point, then adjust based on what you know about the day.

Forgetting to upload sales data

Without sales data, the forecast falls back to burn-rate-style estimates. The "switch" between modes becomes cosmetic. Upload at least 30 days of sales for forecasts to be meaningful.

Related