A Score Out of a Hundred, and What It Weighs
The asset panel shows a health score out of a hundred and a confidence figure beside it. Both are arithmetic you can check by hand: five deductions with named caps, and a confidence that counts how much evidence the score had. Here is every weight in it.
A number between nought and a hundred, on a dashboard, with no working shown, is a number you should not act on. So here is the working.
The asset panel starts every workspace at a hundred and takes points off for five specific conditions. Each deduction has a per-item weight and a cap, so no single condition can dominate the score, and the whole thing is deterministic — the same register on the same day produces the same number every time.
| Condition | Cost per item | Capped at | What counts |
|---|---|---|---|
| Lost or damaged | 5 points | 25 | Assets whose status is lost or damaged. Five of them exhausts the cap. |
| In maintenance | 2 points | 15 | Assets currently in maintenance. Not a fault — but a fleet with a third of itself in the workshop is not healthy. |
| Overdue returns | 3 points | 15 | Checked out, with an expected return date now in the past. |
| Warranty expiring | 1.5 points | 10 | A warranty date falling within the next sixty days. |
| Unverified | proportional | 15 | Never verified, or last verified more than a hundred and eighty days ago — scaled by what fraction of the register they are. |
Retired assets are excluded from every one of those populations. That is worth stating plainly, because it means retiring an asset properly is the one action that both tidies the register and stops it counting against you — which is the correct incentive.
The fifth deduction is the only proportional one
The first four are counts. The fifth is a ratio: the proportion of your register that has not been verified recently, multiplied out and capped at fifteen points. A register where three quarters of the assets have not been laid eyes on in six months takes the full fifteen. A register of two thousand assets with fifty unverified takes less than half a point.
The reason for the difference is scale. Five lost assets is bad in a register of twenty and bad in a register of two thousand, so a count is right. Fifty unverified assets is a stocktake overdue in the first case and a rounding error in the second, so a ratio is right. Mixing the two shapes in one score is deliberate rather than sloppy.
A worked score
Every figure above is readable off the same panel, in the metrics beside the score. The arithmetic is reproducible on paper, which is the property that makes a score like this usable at all.
The floor is twenty, not zero
Add the five caps together and you get eighty. So a register in the worst state the five conditions can describe scores twenty, not nought. The clamp to zero exists as a guard rather than as a reachable outcome. This is a real property of the number and worth knowing before you set a target against it: the useful range is twenty to a hundred, and most working registers sit between fifty and ninety.
Fifty means "not computed", not "average"
When the panel has nothing to work with — a workspace with no assets registered, or a calculation that could not complete — it returns a score of fifty with a confidence of fifteen and no metrics at all. That is a deliberate signal rather than a middling verdict. If you see fifty with an empty metric row and a confidence in the teens, the number is telling you it has no opinion, not that your register is mediocre.
The confidence figure is about evidence, not accuracy
Beside the score sits a second number, and it answers a different question: how much did the score have to go on? It starts at thirty, adds two points for every asset in the register up to a maximum of forty-five, adds fifteen if any purchase costs are recorded at all, and adds one point for each asset currently out up to ten.
- Thirty is the floor. A workspace with a single asset still gets a confidence of thirty-two, which is the panel saying it is guessing.
- Twenty-three assets exhausts the size term. Past that, more assets do not raise confidence, because the score is already reading a real population.
- Recording purchase costs is worth fifteen points on its own — the single largest jump available, and a fair one, since a register with no costs cannot say anything about value.
- Assets currently out add up to ten, on the reasoning that a register nothing ever leaves is a list rather than an operational record.
A high score with a low confidence is a register nobody has tested yet.
The recommendation is a ladder, and it stops at the first rung
Under the score sits one recommendation, chosen by working down a fixed priority order and stopping at the first condition that applies: overdue returns first, then lost or damaged, then warranties expiring within sixty days, then assets unverified for a hundred and eighty days, and finally — if none of those apply — a note that the register looks healthy and to keep verifications current.
This means the recommendation is not the biggest problem, it is the highest-priority category with anything in it. One overdue return outranks eighty unverified assets. That is defensible as a design — chasing an overdue asset is an action somebody can take today, and a stocktake is a project — but it is worth knowing when the panel keeps telling you the same thing.
Five questions to ask about any health score
What are the weights?
A good answer sounds like
A table, published.
What ours actually is
Five deductions: 5, 2, 3 and 1.5 points per item with caps of 25, 15, 15 and 10, plus a proportional term capped at 15.
Can I reproduce it by hand?
A good answer sounds like
Yes, from figures on the same screen.
What ours actually is
Yes. Every input is shown as a metric beside the score.
Is it calibrated against outcomes?
A good answer sounds like
An honest no, or the study.
What ours actually is
An honest no. It is a stated heuristic that encodes a reasonable view of what a well-kept register looks like, not a model fitted to anything.
What does the lowest possible score mean?
A good answer sounds like
A number, and what reaches it.
What ours actually is
Twenty, from the five caps summing to eighty. Zero is a defensive clamp rather than an outcome.
What does the confidence figure measure?
A good answer sounds like
Evidence, not correctness.
What ours actually is
Register size, whether costs are recorded and how many assets are out. It never claims the score is right, only that it had something to read.
Our take
A single number on a dashboard is either a useful summary or an invitation to game it, and what separates the two is whether the arithmetic is published. This one is: five weights, five caps, a proportional term, a clamp, and a confidence figure that says how much evidence there was. Read that way it is genuinely useful — a score that fell twelve points this month has a specific cause you can find in the metrics beside it. Read as a grade, it is misleading in both directions: a hundred means the five conditions are all clear, not that your register is complete or accurate, and nothing in it can see an asset that was never registered in the first place. The score measures the state of what you have told it about, which is the honest limit of every dashboard figure of this kind.
What AWRA OpsHub does today
- A deterministic asset health score from five named deductions, each with a published per-item weight and a cap, clamped to a nought-to-a-hundred range.
- Every input to that score shown as a metric on the same panel — total assets, value, assets out, in maintenance, lost or damaged, warranties expiring, overdue returns and unverified — so the arithmetic can be checked by hand.
- Retired assets excluded from every population in the calculation, so tidying the register improves the score for the right reason.
- A separate confidence figure built from register size, whether purchase costs are recorded and how many assets are currently out.
- A distinct not-computed state — a score of fifty with a confidence of fifteen and no metrics — so an empty or failed calculation is visibly different from a middling one.
- One recommendation chosen from a fixed priority ladder, linking straight to the register filtered for the condition that raised it.
- The same published shape across ten module panels, each with its own weights, so a reader who learns to read one can read all of them.
More we can add to your workspace
- Configurable weights per workspace, so an organisation that considers overdue returns its central risk can say so and have the score reflect it.
- A history of the score over time, with the metric that moved it, so a fall can be attributed rather than investigated.
- A target and an alert on the score, notifying somebody when it crosses a line rather than waiting to be looked at.
- A coverage term rewarding a register where assets carry purchase dates, categories and custodians, rather than only reading the operational conditions.
- A per-department or per-site breakdown of the same score, so a branch dragging the number down is visible without filtering the register by hand.
- A calibration study relating these weights to real outcomes such as loss rates or replacement cost, which would turn a stated heuristic into a measured one.
Where we point you to a specialist
- We will keep publishing the arithmetic rather than hiding it behind a model. A score whose weights are secret cannot be argued with, and an operations figure that cannot be argued with is one people quietly stop believing.
- We will not present this as a benchmark against other organisations. Registers differ enormously in what they contain and how they are maintained, and a comparison across them would be a statement about record-keeping conventions dressed up as a statement about asset management.
- Where an insurer, auditor or funder requires an assessment of your asset controls, that assessment is theirs to make and this figure is not evidence for it. We will produce the underlying registers and movement histories they ask for.
Score history and a per-site breakdown are the contained pieces here — the calculation already runs and its inputs are already stored, so both are a matter of keeping and grouping what is computed.
From a number to a trend
The arithmetic is settled and reproducible. What turns it into an operational tool is remembering it, attributing it and alerting on it.
Score history with attribution
The score kept per day with its metrics, so a fall can be traced to the condition that caused it rather than reconstructed.
A target and an alert
A threshold per workspace, and a notification when the score crosses it, so nobody has to remember to open the panel.
Weights you set
The five deductions exposed as configuration, so the score encodes your view of risk rather than ours.
We publish scope, not dates.
Scope the insight panelsRead the metrics, not the number
The score is a way in. The eight figures beside it are the thing to act on, and they are the same eight a register review would start from anyway.
Talk through asset reportingFrequently asked questions
Is this score generated by a model?
No. It is fixed arithmetic — five subtractions with published weights and caps — computed from counts taken off your own register at the moment the panel is drawn. Nothing is learned, nothing is inferred, and the same register produces the same number every time. The panel sits under an insights heading alongside features that do use models, which is why the question is worth answering plainly.
Why does the score never go below twenty?
Because the five caps sum to eighty and a hundred minus eighty is twenty. The clamp to zero is a guard against a future term being added without its cap being reconsidered, rather than a state any register can reach today. If you are setting a target, the range that matters is twenty to a hundred.
What does "unverified" mean exactly?
An asset that has either never been verified, or whose last verification is more than a hundred and eighty days old. Verification is its own recorded action in the register — somebody confirming an asset is where the system says it is — and it stamps a date and a user onto the record. It is the only one of the five conditions you improve by doing something rather than by fixing something.
Does registering more assets lower my score?
Only through the unverified term, and only temporarily. A newly registered asset has never been verified, so a large import will move the proportion until those assets are verified for the first time. The other four deductions are counts of problems rather than of assets, so growth does not touch them.
Can I see how the score has moved?
Not today — the panel computes the current figure each time it is drawn and does not keep yesterday's. Keeping the score with its metrics per day is the first item on the list above, and it is the change that turns this from a status light into something you can manage against.
Do the other module panels work the same way?
Same shape, different weights. There are ten of them — people, projects, assets, suppliers, point of sale, support, stock, purchasing, sales and accounting — and each starts at a hundred, subtracts named conditions with caps, reports a confidence figure and falls back to fifty when it has nothing to read. Learning to read one teaches you to read all of them, which is the main argument for keeping the shape identical.
