A Practical Data-Informed Manifesto for Small Teams

Illustration of an open book, archival folders, film reel and accordion on a wooden desk.

In brief. Use data to improve a defined decision, document uncertainty and combine measurements with direct evidence from the people affected. This is an original editorial framework.

A data-informed team can explain what it measured, why the measurement matters and how it influenced a decision. This original editorial framework is intended for small teams that need practical habits rather than a dashboard full of numbers. It does not reproduce or claim to restore the wording of another author's manifesto.

Begin with a decision

Write the question before opening the report. For example, a support team might ask whether clearer delivery information reduces repeat enquiries. That question points to relevant evidence: the enquiries, the information customers saw and what happened after a change.

Choose a small set of measures that follow the whole journey. A decline in incoming messages might indicate better information, but it might also mean the contact form stopped working. Check the process directly before declaring success.

Preserve the limits of the evidence

Record the time period, definitions, missing data and any change in how events were counted. Compare like with like. A figure that combines different markets or devices may hide the experience of the people you are trying to help.

Talk to the staff and customers behind the numbers. A short observation can identify a problem that an aggregate measure conceals. Keep that observation attached to its context rather than presenting one person's experience as a universal result.

Make a decision that can be reviewed

Document what the team chose, who is responsible and what would cause a review. If AI helps analyse the material, verify the calculation and the supporting records. NIST's generative-AI profile provides wider guidance on managing the risks of those systems; the small-team examples here are editorial applications.

KINSMC's decision-log guide shows how to keep a traceable record when AI is involved. Start with one decision and a manageable review date. The value of the process is visible when the next person can understand the evidence and revise the choice without reconstructing the entire conversation.

Continue with How to keep a decision log when you use ai at work for related context.

Sources

NIST: Generative AI risk management profile