ReceiptRoller automatically forecasts this month's projected month-end sales and transaction count from past transaction data, as well as demand for your top-selling products up to 14 days ahead. These are displayed as cards on the business dashboard, and can also be retrieved from the Sales Analysis by Product page or via the API/MCP tools.

Types of Forecasts

Monthly Sales and Transaction Count Forecast

The projected month-end sales and transaction count for the current month are calculated with linear extrapolation from actuals to date and the number of days elapsed.

• Example: "This month's sales forecast: ¥5,866,000"
• The card body shows the number of days elapsed (e.g. "26/31 days elapsed, current actual ¥4,919,663") and a confidence interval (¥5,100,000 to ¥6,700,000)
• Transaction count is also forecasted to month end using the same logic

Product Demand Forecast (Up to 14 Days Ahead)

Based on daily sales volume over the past 30 days, demand for the next 14 days is forecasted for your top-selling products (the top 5 by sales volume). The forecast accounts for day-of-week variation (e.g. weekends selling more than weekdays).

• Example: "Drink A: approx. 168 units over the next 14 days (recommended order: approx. 202 units)"
• The recommended order quantity is the forecast value multiplied by a 1.2x safety factor
• Products with less than 7 days of sales history are excluded from forecasting
• Cards are listed under the "Forecast" category on the business dashboard

Determining Confidence Level

The coefficient of variation (CV) is calculated from the last 3 months of snapshots, and confidence is judged on a 3-tier scale.

• High: CV < 10% — month-to-month and day-to-day figures are stable
• Medium: CV < 25% — a typical degree of variation
• Low: CV ≥ 25%, or less than 3 months of history

The confidence interval for the monthly forecast is the forecast value ± CV (capped at 40%). If history is limited, a default of ±15% is applied. Confidence for the product demand forecast is judged from the daily coverage rate (e.g. how many of the last 30 days had sales) and the variability in daily quantity.

Conditions for Running Forecasts

• Monthly Forecast: At least 3 days must have elapsed in the current month, a snapshot for the current month must exist, and sales or transaction count must be 0 or greater
• Product Demand Forecast: There must be sales history for the last 30 days, and the target product must have been sold on 7 or more days

Card Validity Period

The monthly forecast card is valid until the end of the month. It automatically switches to a new forecast for the following month once the month changes. The product demand forecast card is recalculated with the latest 30 days of data whenever the daily scheduler runs at 16:00 (JST).

"This Period's Recap" in Sales Analysis by Product

When the OpenAI integration is enabled, a "This Period's Recap" card is shown above the Sales Analysis by Product page (both the per-store and all-stores versions), summarizing the rule-based analysis cards into a natural single paragraph. Reading it alongside the forecast card lets you grasp the outlook and areas for improvement on a single screen. For details, see Sales Analysis by Product.

Retrieving via API/MCP

Forecast data can be retrieved via the REST API or via MCP tools. Asking an AI assistant (Claude / ChatGPT) "tell me this month's sales forecast" will return this card.

• REST: GET /api/v1/sales/forecast?organizationId={id}&storeId={id}
• MCP: sales_get_forecast

Future Enhancements

• Monthly forecasts at finer daily/weekly granularity
• A non-linear model that accounts for holidays and seasonal factors
• Extending product demand forecasts to 14+ days

Related Articles

• Sales Dashboard
• Sales Trend Analysis
• Sales Analysis by Product
• Customer Loyalty Analysis