# Measure and improve my sales forecast accuracy

An analysis of how accurate past forecasts were, where bias comes from, and the process changes that tighten the number.

- Topic: Data & Analysis
- Framework: Few-Shot
- Level: Standard
- Tags: Sales, Rewrite & Edit, Fill-in Template
- Source: https://www.thinkwithprompt.com/prompts/measure-and-improve-my-sales-forecast-accuracy

## Prompt

```text
Help me measure how accurate our sales forecasts have been and how to improve them.

**Forecast history:** [for each of the last 4-8 periods: forecast at week 1, mid-period, and final week (commit and best case) vs actual closed]
**By rep or team, if available:** [SAME, split]
**Forecast method:** [e.g. rep roll-up, stage-weighted, manager judgment, tool or AI prediction]
**Forecast categories and definitions:** [paste]
**Known issues:** [e.g. late-quarter surprises, sandbagging, happy ears]

Steps:
1. Calculate accuracy for each period and checkpoint: % error and direction (over or under). Show how accuracy improves through the period.
2. Identify bias patterns: consistently high or low, by rep, team, segment, or deal size. Note where a few large deals explain most of the error.
3. Compare commit-category conversion (how much commit actually closed) and best-case conversion to set realistic weightings.
4. Find process causes: unclear category definitions, close dates set by our quarter, large deals not inspected, missing deal-level evidence.
5. Recommend changes and a target accuracy range for each checkpoint.

Return:
- An accuracy table: Period | Checkpoint | Forecast | Actual | Error % | Direction.
- Bias findings by rep or team, stated factually.
- Historical conversion rates by forecast category.
- 5 process changes ranked by impact, with how to measure each.

Work only from the numbers I provide and flag small sample sizes.
```

## What to fill in

- SAME, split
- paste
- e.g. late-quarter surprises, sandbagging, happy ears

Free to copy and use. From ThinkWithPrompt (https://www.thinkwithprompt.com/), a library of 2,663 curated AI prompts.