# What is a baseline measurement, and why does a system need one before it starts?

A baseline measurement is what was true before a system existed: how long a task took, how often it went wrong, how many were handled by hand. Without it, “the system helped” is an opinion. With it, a result can say both ends, the number before and the number after, and say when each was measured.

Also called: baseline
Checked: 2026-09-27

In one line

What was true before the system, so “improved” means something.

## Why it matters to you

The place this gets skipped is a sales conversation, where a percentage sounds definitive without ever naming what it is a percentage of. A Consult exists partly to set this number before anything is built, so year one and year two are compared against a fact instead of a memory.

It also has to be dated and system-level: an hours figure and a date it was measured, not a projection and not a channel metric standing in for the whole system. A number without both ends, before and after, is a claim nobody can check.

## What it is not

A baseline measurement is not a market average or an industry benchmark. It is the buyer's own number, taken from the buyer's own records, before their system existed.

## Next to this

- [Total cost of ownership](https://octyn.co/glossary/total-cost-of-ownership)
- [p95 latency](https://octyn.co/glossary/p95-latency)
- [Rollback](https://octyn.co/glossary/rollback)

## Argued out at length

- [What it costs to change an AI workflow six months later](https://octyn.co/compare/cost-to-change-ai-workflow-six-months)

## Related

- [What it costs to change an AI workflow six months later](https://octyn.co/compare/cost-to-change-ai-workflow-six-months)
- [Total cost of ownership](https://octyn.co/glossary/total-cost-of-ownership)
- [The Consult](https://octyn.co/consult)

Canonical: https://octyn.co/glossary/baseline-measurement