Ivara Labs
Dashboards fail quietly. Nobody announces that they have stopped looking. The screen stays up on a wall or a browser tab, the numbers keep refreshing, and at some point the team goes back to asking each other questions instead.
The usual explanation is the wrong tool. It rarely is. The dashboard failed because it was assembled from the metrics that were easiest to extract, rather than the ones that change a decision.
The test every metric has to pass
If this number moved sharply tomorrow, who would see it, and what would they do differently that same week?
If you cannot name the person and the action, the metric is decoration. It may be interesting. It is not operational, and it is competing for attention with something that is.
Run this against an existing dashboard and most tiles fail. Total page views. Number of active clients. Tasks completed this month. All true, all uncontroversial, none of them prompting anybody to do anything today.
The tiles that survive tend to be the ones a person is already anxious about and currently checks by asking somebody.
The four numbers a service business runs on
For an operation that sells people's time and expertise, four numbers carry most of the weight. Everything else is either a component of one of them or a detail you can pull when you need it.
Each one has a decision attached. That is the qualification, not the fact that it is measurable.
Operations running across several sites usually want a fifth: locations per coordinator, which shows whether growth is compounding or simply accumulating overhead.
Margin by client is the one most operations lack and most need. Revenue by client is easy and mildly misleading — the largest account is frequently not the most profitable one, and without hours against it nobody finds out until somebody leaves and the numbers improve.
Cash position carries a similar dependency. It is only as current as the ledger behind it, which is the argument for reconciling daily rather than at month end — a dashboard reading from books closed seven weeks ago is reporting history.
Most dashboards are entirely rear-view
The second common failure is that every number describes the past. Last month's revenue, jobs completed, invoices sent. All lagging, all arriving too late to change.
A useful dashboard pairs each lagging number with the leading indicator that moves before it.
| Lagging — tells you what happened | Leading — tells you what is coming |
|---|---|
| Revenue last month | Proposals out, work committed |
| Margin at quarter end | Hours logged against estimate, live |
| Churn last year | Response times, overdue items per client |
| Invoices paid | Ageing receivables by bracket |
| Jobs delivered late | Jobs with no owner or no next date |
The right-hand column is where intervention is still possible. A job that has no owner today becomes a late delivery in three weeks, and that is a problem you can solve now for almost nothing.
One number set, one definition
This is the least glamorous point and the most common cause of a dashboard nobody trusts.
Ask three people in the same company what revenue means and you can get three answers: booked, invoiced, or collected. Utilization might mean billable hours over available hours, or over contracted hours, or over paid hours including leave. None of these is wrong. All of them being live at once is fatal, because every meeting starts by relitigating whose number is right.
A number nobody trusts is worse than no number. It costs the same to produce and generates an argument instead of a decision.
Write the definitions down, put them next to the metric, and make sure every view reads from the same source. Client-facing reporting and internal reporting should be the same data with different framing, never two separate builds that drift apart within a quarter.
Alerts, not just charts
A dashboard you have to remember to open is a report with better styling. The genuinely useful version tells you when something has moved, without being asked.
Pick the handful of conditions that matter and route them where people already are — a message, not a page. Margin on an active job crossing below target. Hours logged passing eighty percent of estimate. A receivable crossing sixty days. A job with no activity for a week. Cash projected below a floor within thirty days.
Keep the list short. Alerts that fire constantly get muted, and a muted alert is worse than none because everyone believes it is still watching.
What to take off the wall
Tiles that usually fail the so-what test
- Counts with no denominator. Forty jobs completed is meaningless without capacity, target or last period beside it.
- Revenue by client without hours. Flattering, and it hides the accounts quietly costing you money.
- Anything rebuilt by hand. If a number is assembled manually each month, it is stale on arrival and nobody will act on it in week three.
- Metrics with no owner. If nobody's job is affected by the number, nobody will notice when it moves.
- Vanity totals. Cumulative all-time figures only ever go up. They cannot tell you anything.
Before you build one
Two questions decide whether a dashboard is even the right project.
Does the data exist where the work happens? A dashboard reading from the live work stays true. One assembled from exports is a report on a schedule, and it will drift. If the underlying systems do not hold the data, that gap is the actual project.
Is anyone accountable for each number? Metrics without owners are observations. The dashboard is not what changes behaviour — the named owner is, and the dashboard is how they see it early.
How we build these
We model the operation rather than dropping in a template, so cash, pipeline, margin and utilization come from the work itself and nothing goes stale. One number set, one definition, with client and internal views reading the same live data.
And the system flags margin, hours or cash drifting off pattern before the month closes, rather than waiting to be asked. That is what a command center is, and it is one of the four layers running inside the platform we operate our own company on.
If your numbers currently live in five exports and three opinions, the useful starting point is a map of how the work actually moves.
Frequently asked questions
- What metrics should a service business dashboard show?
- Four carry most of the weight: cash position, pipeline, margin by job or client, and utilization. Each has a decision attached to it. Everything else is usually a component of one of these or a detail to pull on demand.
- How do I know if a metric belongs on the dashboard?
- Apply the so-what test: if the number moved sharply tomorrow, who would see it and what would they do differently that week? If you cannot name both the person and the action, it is decoration.
- What is the difference between leading and lagging metrics?
- Lagging metrics describe what already happened, such as last month's revenue. Leading metrics move first and leave room to intervene, such as hours logged against estimate on live jobs. Most dashboards are almost entirely lagging.
- Why does nobody use our dashboard?
- Usually one of three reasons: the metrics fail the so-what test, the definitions are contested so nobody trusts the numbers, or the data is assembled by hand and is stale by the time anyone looks.
- Should a dashboard send alerts?
- Yes, for a short list of conditions that need action — margin crossing a threshold, hours passing a share of estimate, receivables ageing past a bracket. Keep the list small, because alerts that fire constantly get muted.