> ## 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.

# Forecast accuracy & Base Price Performance

> Track how accurate your demand forecast has been week to week, and measure the realized benefit of the price changes you already made.

The [sales forecast](/agents/sales-forecast) tells you what a price change is *expected* to do before you approve it. **Forecast accuracy** and **Base Price Performance** are the other half of the loop: they look *backward*, at how the forecast actually held up and at what your past price changes really earned. Together they let you trust — and tune — the numbers you price on.

<Note>
  This page covers the backward-looking view. For the forward-looking impact shown at approval time, see [Sales forecast & price impact](/agents/sales-forecast).
</Note>

## Why look backward

A forecast you never check is a guess you keep repeating. Two questions decide whether to keep trusting it:

* **Has the forecast been right?** If last week's forecast said 120 units and you sold 60, the model is telling you something — about that item, that category, or that season — worth knowing before you price on it again.
* **Did the price change pay off?** A change looked good in the forecast at approval time. Weeks later, with real sales in, you can see whether it actually moved gross profit the way you expected, or whether the market moved instead.

## Forecast accuracy

Forecast accuracy compares each week's forecast against what actually sold once the week closed. Read it to find where the model is dependable and where it is not:

* **Where accuracy is high**, the forecast is a reliable input — you can lean on the approval-time impact numbers and, where you choose to, let scheduled runs act on them.
* **Where accuracy is low**, treat the forecast as a hint rather than a number. Thin sales history, a promo the model didn't see, or a genuine demand shift all show up here as a gap between forecast and actual.

Accuracy is tracked per week, so you see a trend rather than a single snapshot — a model that is drifting looks different from one that missed a single unusual week.

## Base Price Performance

Base Price Performance (BPP) measures the **realized** effect of the price changes you already made — the difference between how an item sold after a change and how the model expected it to sell at the old price over the same period. It answers "did that repricing actually earn what we thought it would?" in money, not in theory.

Use it to:

* **Settle whether a pricing pass worked.** Instead of arguing from the approval-time forecast, read the measured benefit once real sales are in.
* **Learn which moves pay off.** Patterns across items — which categories, roles, or kinds of change delivered — feed directly into how you set up the next run.
* **Catch changes that quietly cost you.** A move that looked fine at approval but underperformed is exactly what BPP surfaces.

## How the two fit together

Forecast accuracy tells you how much to trust the model; Base Price Performance tells you whether acting on it paid. Read accuracy first — if the forecast has been reliable for a segment, its BPP is a clean read on your pricing decisions; if accuracy is poor there, a weak BPP is as likely a model problem as a pricing one, and the fix is better data or more history, not a different price.

## Related

* [Sales forecast & price impact](/agents/sales-forecast)
* [Price Optimization](/agents/price-optimization)
* [Review and approve price changes](/agents/price-approval)
* [Metrics glossary](/reference/metrics)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.