Operational Dashboards: Choosing the Six Numbers That Matter
A dashboard with twenty tiles is a wall decoration. Choosing the six numbers that would have warned you about last year's surprises — and the rule that keeps a dashboard useful once it exists.
Dashboards fail in a specific and consistent way. Somebody builds a good one, everybody looks at it for two weeks, and then it becomes furniture — present, glanced at, no longer read. The failure is not the design. It is that most tiles on it never change in a way that requires anybody to do anything.
A number worth watching has a threshold beyond which somebody acts. A number without one is context, and context belongs in a report you open when you have a question, not on a screen competing for daily attention.
Choose from your surprises, not from a template
The best method for picking dashboard numbers has nothing to do with best-practice KPI lists. Sit down and write out everything that genuinely caught your business out in the past twelve months.
A stockout on a line that sells well. A customer who had quietly stopped ordering. A project that came in below margin. A month where cash was tight and nobody flagged it. A supplier price increase absorbed for a quarter before anyone noticed. Each of those surprises corresponds to a number that was not being watched — and that list, rather than a generic KPI framework, is your dashboard.
Surprise is the signal. Every unpleasant surprise in the last year is a number nobody was watching, and it is a far better guide to what belongs on a dashboard than any best-practice list.
Every tile needs four things
Before a number earns a place, it should have all four. Missing any one is why tiles become decoration.
| Requirement | Why | The test |
|---|---|---|
| An owner | A number nobody owns is nobody's problem | Can you name the person who acts on it? |
| A threshold | Without one, every value looks acceptable | At what value does somebody do something? |
| An action | The threshold has to trigger a specific response | What exactly happens when it is breached? |
| A rhythm | It must be looked at often enough to be actionable | Daily, weekly or monthly — and does it match the decision? |
The threshold requirement does most of the filtering. "Total sales this month" has no threshold — it is always just a number, and it is on almost every dashboard in Kenya. "Receivables over 60 days above 2 million" has one, and when it breaches somebody makes calls. One of those changes behaviour.
A worked example set
Not a template to copy — an illustration of the shape. Six numbers, each with an owner, a threshold and a rhythm.
What a working six looks like
- Stock cover on A-class items — ops, weekly. Threshold: any A item under two weeks' cover. Action: raise a purchase requisition. See ABC analysis.
- Receivables over 60 days — finance, weekly. Threshold: a set shilling value or any single account over a limit. Action: work the named invoices, then credit hold.
- Committed vs available on live projects — project lead, weekly. Threshold: any line under 20% available. Action: stop further commitment and review.
- Margin by customer, bottom five — sales, monthly. Threshold: any account below a floor margin. Action: renegotiate terms or accept it deliberately.
- Tickets by category — support, monthly. Threshold: any category above its normal range. Action: fix the cause rather than staff the symptom.
- Customers who have gone quiet — sales, monthly. Threshold: any regular buyer with no order in an unusual interval. Action: a phone call this week.
Notice what is absent: revenue, total stock value, headcount, ticket count. All useful context, none of them decision triggers. They belong in the monthly pack, not on a screen somebody is supposed to scan.
Match the rhythm to the decision, not to the accounting calendar
The most common dashboard error in Kenyan businesses is putting everything on a monthly cycle because that is when finance closes. But the decisions have their own natural frequencies, and a mismatch makes an accurate number useless.
Monthly for everything
- A stockout forms and completes inside one reporting period
- A project overruns before anybody sees committed cost
- A cash squeeze is confirmed after it has happened
- A customer has fully drifted before appearing on a list
- Every review is a post-mortem, so nobody feels ownership
Matched to the decision
- Stock cover weekly, because that is how fast it moves
- Committed cost weekly, while a commitment can still be stopped
- Receivables weekly, so a squeeze is three weeks of warning
- Quiet customers monthly, which is soon enough to recover them
- Reviews produce actions, so the rhythm sustains itself
Push it, do not host it
Even a well-designed dashboard requires someone to log in and look, which happens reliably when things are calm and stops exactly when things are busy — the moment it matters most. Pinning the six numbers together is useful; scheduling them to the owners is what makes them survive a hard quarter.
So do both: pin them for the person who wants to look, and schedule them for the person who will not. The delivery side is covered in scheduled reports.
Retire tiles deliberately
Once a quarter, ask of each tile: has anybody acted on this in the last three months? If not, either the threshold is wrong or the number does not belong. Removing a tile is a decision, not an admission — a dashboard that only ever grows becomes unreadable within a year, and then the two numbers that mattered get lost among eighteen that did not.
What we do and do not do
What AWRA OpsHub does today
- Dashboard pinning, so your chosen numbers sit together rather than being hunted through a menu.
- Saved report definitions and filters, so a view is kept rather than rebuilt.
- Scheduled delivery to named users or whole roles, on a cadence.
- Reports from live operational records, so the number reflects the transaction rather than a transcription.
- Sharing, so a dashboard has an audience rather than an owner.
What it does not do
- No threshold-based alerting you configure per tile. The thresholds in this guide are a management discipline — you set them and act on them; the system does not page you when a tile breaches.
- No drag-and-drop dashboard designer with arbitrary chart types and layout control.
- No external data sources blended in alongside operational data.
- No goal or target tracking with variance colouring against a plan you upload.
That first point matters for expectations: pinning and scheduling get the numbers in front of the right people on a rhythm, but the threshold and the action stay a human discipline. Do not build a dashboard strategy that assumes the system will interrupt you.
The failure to watch for
A dashboard that everybody agrees is excellent and nobody has acted on is worse than none, because it provides the feeling of control without the substance. The diagnostic is simple and slightly uncomfortable: at the next management meeting, ask what decision was made in the last month because of something on the dashboard.
If the answer is nothing, the dashboard is not underused — it is wrongly specified. Go back to the surprises list, check that each number has an owner, a threshold and an action, and cut whatever fails. Six numbers people act on beats twenty everybody admires.
Our take
Build your dashboard from last year's unpleasant surprises, not from a KPI template. Give every tile an owner, a threshold and an action, match the rhythm to the decision rather than to month-end, and retire anything nobody has acted on in a quarter. Then schedule it, because the person who most needs the number is the one least likely to log in and look for it.
See the numbers you actually act on
Pin your chosen reports to a dashboard, save the definitions and filters behind them, and schedule delivery to the people who own each number.
Explore dashboards & reportsFrequently asked questions
How do we choose which numbers to watch?
List everything that genuinely caught your business out in the last twelve months — a stockout, a customer who quietly stopped ordering, a project that lost money, a tight month nobody flagged. Each surprise corresponds to a number that was not being watched, and that list is a far better guide than any best-practice KPI framework, because it is specific to how your business actually fails.
Why is total revenue a bad dashboard tile?
Because it has no threshold. It is always just a number, so no value triggers an action, which means looking at it changes nothing. Compare it with "receivables over 60 days above a set value" — that breaches, and when it does somebody makes calls. Revenue is useful context for a monthly pack; it is not a decision trigger, and dashboards should hold only decision triggers.
Does the system alert us when a number crosses a threshold?
Not as per-tile threshold alerting you configure yourself, and it is worth being clear about that. You can pin reports to a dashboard and schedule them to the people who own each number on a cadence, but the threshold and the response stay a human discipline. Do not design a monitoring strategy that assumes the system will interrupt you when something breaches.
How often should a dashboard be reviewed?
Match each number to its decision rather than putting everything on the accounting calendar. Stock cover and committed project cost move fast enough to need weekly attention; margin by customer and quiet-customer lists are monthly. Putting everything on a monthly cycle because that is when finance closes means a stockout can form and complete inside one reporting period.
Our dashboard looks good but nothing changes. What now?
Ask at your next management meeting what decision was made in the last month because of something on it. If the answer is nothing, the dashboard is wrongly specified rather than underused. Go back through each tile and check it has a named owner, a threshold value and a defined action; cut whatever fails all three. Six numbers people act on is worth more than twenty everybody admires.
