Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.retailgrid.io/llms.txt

Use this file to discover all available pages before exploring further.

Sales Forecast estimates future sales per SKU using your historical transaction data. Use it for demand planning, seasonality analysis, capacity decisions, or as input to other agents that benefit from forward-looking volume estimates.

​
When to use Sales Forecast

Reach for Sales Forecast when:
  • You’re planning purchase orders or production runs and need a per-SKU volume estimate.
  • You’re analyzing seasonality and want a baseline curve to compare against.
  • You want to feed forecasted volumes into Price Optimization or a custom margin model.

​
Open Sales Forecast

In any grid:
  1. Click Agents in the top toolbar.
  2. Select Sales Forecast from the All Agents modal (Analytics Agents tab).
  3. The configuration panel opens with three tabs: Config by AI, Columns, Parameters.

​
Configure in three tabs

​
Columns - map Sales history

  • Sales / Transactions - the historical transaction data the model reads.
  • Date column - the timestamp on each transaction (defaults to transaction_date).
  • Quantity column - how many units were sold per transaction (defaults to quantity).
Without transaction data the model has nothing to forecast.

​
Parameters - set the horizon and granularity

  • Forecast horizon - how far ahead to forecast (7 days, 30 days, 90 days, or custom).
  • Granularity - daily, weekly, or monthly buckets.
  • Seasonality - opt in to weekly and yearly seasonality components if your category shows them.
  • History window - how far back to look (defaults to 365 days).
Defaults assume monthly granularity over a 90-day horizon. For fast-moving SKUs (groceries) consider weekly granularity.

​
Config by AI

Describe what you want in plain language - example: “Forecast next quarter for the Beverages category, daily granularity, capture weekly seasonality.”

​
Read the result

After a run, the grid shows new columns:
  • Forecast Units (next period) - estimated units in the configured horizon.
  • Forecast Confidence - prediction interval width (narrower = higher confidence).
  • Trend - directional indicator (rising / flat / declining vs. recent history).
Click a forecast cell to open the right-side Price Analysis panel - same panel as the other agents, here showing the forecast trace, seasonality components, and the historical fit quality.

​
Common pitfalls

  • Sparse history - SKUs with fewer than ~30 transactions get a flat forecast. The model doesn’t fabricate signal.
  • Ignoring confidence - the point estimate is what most users see, but the confidence interval matters for downstream planning. Wide intervals = act on the bound that fits your risk tolerance.
  • Pairing with Price Optimization - if you change prices via Price Optimization, the historical elasticity assumed by Sales Forecast may shift. Re-run Sales Forecast after major price moves.