Most organisations do not have a data problem. They have a decisions problem dressed as a data problem. Dashboards proliferate, numbers multiply, exports are emailed each morning — and the team is no closer to knowing what to do next.
Good operational data is quieter than that. It exists to support a small number of recurring decisions, and to make exceptions impossible to ignore.
Three things worth getting right
1. A single source of truth, per concept
For each important entity — customer, job, booking, vehicle, product — there should be exactly one place that is considered authoritative. Other systems can hold copies, but the rule is clear: when they disagree, this one wins. Without that rule, every report is a negotiation.
2. Decisions, not vanity
A useful dashboard answers a question someone will act on. "How many enquiries this week?" is rarely actionable on its own. "How many enquiries this week are still waiting on a response after 24 hours?" almost always is. The first is a metric; the second is a decision.
3. Exceptions surfaced, normality hidden
Operational data should make the unusual loud and the routine quiet. A well-designed view shows you what needs attention, not everything that happened. The team's time is the scarce resource, not the data.
Things to be sceptical of
- Dashboards that nobody opened last week.
- Reports that always look healthy, regardless of the week.
- Numbers without a definition anyone can recite.
- "More data" as the answer to "why didn't we know?"
Where to start
Pick one decision that recurs every week. Write down what data would make it obvious. Make that data available to the person who has to make the decision. Do nothing else until that works. Most data programmes fail because they begin with the warehouse rather than with a single recurring choice.