Priced per environment, not per incident

The obvious model would be per investigation. It sounds fair — you pay for what you use — and it is exactly wrong, because it puts a price on the moment someone is deciding whether a number looks odd enough to look into. At the margin, a per-investigation price does not produce revenue. It produces a shrug and a guess.

That matters more here than in most categories, because the failures this product is built for are the quiet ones. Nobody hesitates to investigate a pipeline that crashed. The 38% shortfall on a run that reported success is caught by somebody who thought a figure looked slightly off and decided to check — precisely the decision a per-use price discourages.

Per seat has the same defect in a different place. An investigation gets better when the people who know the pipeline can challenge a conclusion, attach the export nobody else has, or run the read-only check Decim wrote for them. Charging per head shrinks that group deliberately.

So the unit is the environment: a thing that exists whether or not anything went wrong this month, and whose value does not change with how carefully you use it.

Everything that makes an investigation trustworthy is in every plan, including the free one: pipeline discovery, cited log lines, competing hypotheses with the evidence for and against, checks for your team to run, draft fixes and the report. Those are not premium features; they are the product. What varies is scale and support. Rows marked planned are not available on any plan yet.

An environment is a set of systems you would investigate as one thing. Usually that maps to production, staging, and sometimes a disaster-recovery estate. The test is practical rather than architectural: if an incident in one place would never send you looking at the other, they are separate environments.

What an environment explicitly does not depend on:

• How many databases, services or repositories it contains
• How many pipelines run in it
• How many log sources you connect, or agents you install
• How many people look at the results
• How many incidents happen in it

A single production environment with forty services in Python, .NET and Java, a set of scheduled SQL jobs, a few cloud ETL jobs and a warehouse is one environment. That is deliberate: a model that charged per component would penalise exactly the teams with the most to investigate.

Each of these is a case where a meter would change behaviour in a direction that damages the thing you are paying for.

Related: design partners for what a pilot involves; security for what Decim reads, where evidence goes and how long it is kept; and the product for what is actually included.