> ## Documentation Index
> Fetch the complete documentation index at: https://docs.spherecast.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Forecast accuracy

> Compare past forecast snapshots against actual sales to see how good your plan was

## What this page is for

Forecast accuracy answers a simple question: how close was the forecast to what actually sold? Spherecast compares a past **snapshot** of your plan against actual sales and scores the miss. You can measure any metric on the ladder, and you can control how far in advance the forecast was made.

> This may be a feature some companies don't have turned on. If you don't see it, it isn't enabled for your company.

## What you need first

Accuracy needs **snapshots** — frozen copies of the plan taken in the past. If none exist yet, there's nothing to compare against. Create snapshots from the demand plan (see **[Building consensus](/guide/demand-planning/building-consensus)**).

## The accuracy metrics

| Metric          | Full name                      | What it tells you                                            |
| --------------- | ------------------------------ | ------------------------------------------------------------ |
| **MAE**         | Mean Absolute Error            | The average size of the miss, in units                       |
| **MAPE**        | Mean Absolute Percentage Error | The average miss as a percentage                             |
| **Bias (Abs.)** | Absolute bias                  | Whether you systematically over- or under-forecast, in units |
| **Bias %**      | Percentage bias                | The same systematic lean, as a percentage                    |

MAE and MAPE tell you **how far off** you were. Bias tells you **which direction** you consistently lean — a positive bias means you keep forecasting too high, a negative bias too low.

Each metric can be measured for **Target**, **Baseline**, **Baseline Adj.**, or **Consensus**, so you can see which row of the ladder was most reliable.

## Lag: how far ahead the forecast was made

A **lag** is the number of months between when the forecast was made and the month it was forecasting.

* **Lag 1** = the forecast made one month ahead.
* **Lag 3** = the forecast made three months ahead.

You pick which snapshots to use by lag. **Unavailable lags are greyed out** — you can only compare lags for which a snapshot exists.

### Compare lags to month

This choice controls how each month lines up with its snapshot:

| Setting     | What it does                                                                  |
| ----------- | ----------------------------------------------------------------------------- |
| **Current** | Every month uses the snapshot from the same number of months ago              |
| **Each**    | Each month shows the forecast that was made that many months in advance of it |

## Plotting over time

You can also chart snapshot time-series — for example **Baseline (Snapshot)** and **Consensus (Snapshot)** — against **actuals**, so you can see visually where the plan drifted.

## Step by step: check last quarter's accuracy

1. Confirm snapshots exist for the period you want to review.
2. Open forecast accuracy and choose the metric — for example **MAPE**.
3. Choose the ladder row to score — for example **Consensus**.
4. Pick the **lag** (grey lags have no snapshot).
5. Set **Compare lags to month** to **Current** or **Each**.
6. Read the score, and optionally plot the snapshot series against actuals.

## Example

A planner checks Consensus at Lag 1 for last quarter. MAPE comes back at 12%, and Bias % is +4%, meaning the plan was usually a little high. At Lag 3 the MAPE rises to 21% — the further ahead they forecast, the less accurate they were. Plotting Consensus (Snapshot) against actuals shows the gap widened during a promotion month.

> **Tip:** Accuracy only works if you take snapshots consistently — ideally every planning cycle. A missed snapshot means a greyed-out lag you can never recover. See **[Building consensus](/guide/demand-planning/building-consensus)** to make snapshots a habit.