How to choose company health metrics that guide action

A manager reading a long report by the light of a desk lamp

Choose company health metrics by listing the decisions you need to make, then selecting a measure that can inform each one. For every measure, record what action it could change and what it cannot tell you; a metric is useful only in the context of the decision it serves.

Which decisions should your metrics help you make?

A dashboard can collect measures without helping anyone choose what to do. Start instead with decisions you expect to face soon. These might include whether to change a price, slow hiring, focus on a delayed project, or spend more on a sales channel.

For each decision, write down the choice the evidence could help distinguish. If you are deciding whether to continue a channel, for example, you may need to know whether it brings customers who stay and generate enough revenue. A count of new leads may show activity, but it may not help distinguish a healthy channel from an expensive one.

Separate a decision trigger from background context. A trigger is evidence that could change the choice. Context helps explain the situation, but may not affect what you do. Both can be useful, but they should not be presented as if they serve the same purpose.

Drop measures that do not connect to a decision. If nobody can name an action that might change when the measure changes, it may be a report rather than a decision tool. That does not make it worthless. It does mean it should not take attention away from measures tied to choices you need to make.

How do you choose company health metrics?

For each candidate, write down the decision it serves and the action it could change. Then name the source and check whether it is current enough for that decision. A measure based on delayed or partial records can still help, but its limits need to be visible before you act on it.

Next, distinguish observed results from expectations. A result describes what the records show. An estimate or forecast describes what may happen, based on assumptions. Keep those separate. Otherwise, a team can mistake a plan for evidence that the plan is working.

Prefer a short list with distinct jobs. Two measures that answer the same question may add clutter rather than clarity. A useful set might include a measure that shows demand, another that shows whether work is being delivered, and one that helps assess the economics of serving customers. The right set depends on the decisions in front of you, not on a standard dashboard.

Brainis maintains a live model of company state with sources, freshness, and confidence. That gives you a shared basis for checking where a measure came from and how current it is. Read about company state in Brainis, or see what company state means.

Example: A fictional small design studio is deciding whether to keep offering a fixed-price setup service. Its owner tracks the hours spent on each setup and the fee collected, using project records as the source. Those measures could change whether the studio keeps the offer, changes its scope, or raises its price; they do not, by themselves, explain why some jobs take longer.

What can a metric not tell you?

A measure can describe an outcome without identifying its cause. If revenue falls, the number alone does not say whether fewer customers bought, customers spent less, delivery slowed, or several things changed together. Treat the measure as a signal to investigate, not a complete explanation.

A company-wide result can also hide differences between customers, teams, or features. The total may look acceptable while one part of the business loses money or another part carries the result. Where the decision depends on those differences, look for a measure at the level where the action will be taken. Brainis joins model-provider usage and revenue into per-customer, per-feature margin, with lineage for each cent and negative-margin customers flagged with a proposed fix.

Forecasts need their assumptions and ranges beside them. A forecast is not an observed result, and a single expected value can conceal how uncertain the outcome is. Brainis models alternative futures before recommending a path. Compare those futures with the evidence you have, and record what new evidence would change how you interpret the signal. The world model explains that Brainis capability; why ranges beat certainty discusses how to keep uncertainty visible.

Two colleagues tracing connections between notes pinned on a glass wall

How should you review the measures?

Review the decision, measure, source, and resulting action together. Ask what the team expected to happen, what actually happened, and whether the measure helped someone choose. Keep the original expectation intact. If you rewrite it after seeing the result, you lose the comparison that would show what the team learned.

When a measure did not change the action, ask why. Perhaps the source was too old, the measure did not distinguish between the available choices, or the decision had already been made for another reason. Fix the gap if the decision still matters. Retire the measure if it no longer informs a choice.

This review is a learning loop, not a test of whether a person was right. A forecast can miss because the assumptions were weak, because conditions changed, or because the evidence was incomplete. Write down which explanation fits the available evidence, and what would help tell them apart next time.

Brainis records predicted and actual outcomes in the Decision Ledger and uses them to improve the next recommendation. You can read about memory in Brainis and about writing predictions down. The practical test remains whether the recorded outcome helps you make a better-informed decision later.

What should you do next?

Pick one upcoming decision and write it as a question. Choose one candidate measure that could change the answer, and name the record or system it comes from. If the source is unclear, resolve that before relying on the measure.

Write down what the measure cannot establish. It may show a change without explaining its cause. It may combine customers with different economics. It may describe what has happened while leaving the next outcome uncertain. Naming the limit makes it less likely that a useful signal will be mistaken for the whole story.

After the decision, record the action and the result you expected. When the result is available, compare it with that expectation without changing the original. Ask whether the measure changed the action, and whether the result changed what you would look at next time.

That is enough to begin choosing company health metrics with purpose. Add another measure only when it answers a distinct decision question. Keep the source, limits, and action close to the measure, so the team can see what it supports and what remains unknown.