What sets the figure
The five conditions that size a predictive maintenance budget
None of these is a number, and each can be examined before any supplier is invited in. A quote produced without them has assumed a plant that is not yours.
| Variable | Effect on price | Why |
|---|---|---|
| What the plant already measures | The largest single swing in the programme | A facility whose critical equipment already carries vibration, temperature and current sensors feeding a historian starts from data; a green-field plant starts from procurement, installation and cabling in an operating environment. The modelling work is similar in both — the instrumentation gap is what separates their budgets, and it is visible before anyone quotes. |
| Depth and quality of failure history | Decides what the models can learn, and how soon | Models learn failure from examples of failure. Years of labelled work orders and clean historian data shorten the path to a trustworthy prediction; a history of hand-written logs, or of maintenance so preventive that failures were never allowed to happen, means the programme spends its first phase building the record it will later learn from. Data archaeology is real work, and it belongs in the plan rather than in the overrun. |
| How many assets, and how critical | Coverage is a choice, not a default | Every asset added brings sensors to read, a model to build and tune, and alerts someone must answer. The honest way to scope coverage is by the cost of each asset's downtime: a handful of machines whose failure stops the line justify the effort long before the hundreds whose failure is an inconvenience. Programmes priced per asset across the whole register are usually paying to model equipment nobody would miss. |
| The path to the CMMS or ERP | Where a prediction becomes a work order — or a dashboard | A prediction that does not raise a work order changes nothing about how the plant is maintained. Connecting the models to your maintenance system — so an alert carries the asset, the evidence and a suggested window into the planner's queue — is integration work whose effort depends on the age and openness of that system. It is often underestimated precisely because it is the least glamorous line in the programme. |
| Model operations after go-live | A running category for as long as predictions are trusted | Equipment is overhauled, operating points move, sensors drift and fail — and a model trained on the old plant quietly degrades against the new one. Monitoring prediction quality, retraining on fresh history and retiring models that no longer earn their alerts is a permanent workload. A budget that ends at go-live has funded the launch of a system and not the years in which it is supposed to pay for itself. |
Ask any supplier quoting this
- Which of our assets does your figure assume are already instrumented, and who pays to close the gap?
- Have you examined our failure history, and what happens to the plan if it turns out too thin to learn from?
- How was the list of covered assets chosen — by downtime cost, or by counting the register?
- Does the figure include the integration that turns a prediction into a work order in our CMMS?
- Who monitors and retrains the models after go-live, and what does that cost as its own line?
Where the money goes
The route a vibration reading travels before it becomes a work order
Each hop on this route is a budget line. A quote that names only the model has priced the middle of the journey and left both ends to be discovered.
A flow from sensing on the asset, through collection and prediction, to a dispatched work order in the CMMS — with a branch where the maintenance engineer can overrule the model.
- 01
Sense
Sensors on the asset capture vibration, temperature or current — bought and installed where they do not yet exist.
- 02
Collect
Readings land in a historian or platform reliably enough to train on and to score against, around the clock.
- 03
Predict
Models score the streams and separate a developing failure from the ordinary moods of rotating equipment.
- 04
Dispatch
The prediction enters the CMMS as a work order with the evidence attached, inside the planner's existing routine.
The engineer overrules
A maintenance engineer can always reject or reschedule what the model proposes. The programme is designed so that trust is earned prediction by prediction — an alert stream nobody believes is the most expensive way to change nothing.
Why no menu
A per-asset price for predictive maintenance flatters the wrong plant
The per-asset figures that circulate in this market assume the convenient plant: instrumented, historised, with an open maintenance system waiting for the integration. Facilities like that exist, and if yours is one, a programme lands quickly. Most are not — and for them a per-asset anchor is worse than useless, because it prices the modelling and silently omits the instrumentation, the data archaeology and the CMMS work that will dominate the real bill.
The alternative is not mystery; it is sequence. An instrumentation and data-readiness assessment measures the actual starting point, the asset criticality review decides what deserves coverage, and only then does a figure describe your programme. That assessment is a bounded, inexpensive piece of work — and it is the difference between budgeting a project and budgeting a hope.
Our role
What we inspect before pricing failure prediction
- 01
The plant as it is
We assess the instrumentation, the historian and the failure records before quoting, so the figure prices the distance from your starting point rather than from a convenient one.
- 02
Coverage argued from downtime
We rank assets by what their failure costs the operation and scope the models to the ones that justify the effort — a shorter list that pays, rather than a long one that impresses.
- 03
A work order, not a dashboard
We build the integration that lands predictions in your CMMS inside the planner's routine, and we set up the model monitoring and retraining that keep the alerts worth answering.
Which failures the plant can tolerate, and which downtime is unacceptable, are judgements only your operation can make; we turn them into a covered-asset list, a data plan and a figure that describes your plant.
Common questions
What buyers ask
Why does no serious supplier print a per-asset price for predictive maintenance?+
Because the asset is not the unit of cost — the plant is. The same pump costs entirely different amounts to cover depending on whether it is already instrumented, whether its failure history was ever recorded, and how hard it is to reach the maintenance system that will act on the prediction. A per-asset price has to assume answers to all three, and whichever plant it assumes is not yours. We would rather spend a short assessment measuring your starting point and quote a programme, than print a number that is precise about someone else's facility.
Our failure records are thin — does that make predictive maintenance unaffordable?+
It changes the shape of the programme more than its feasibility. With thin history the first phase leans on condition monitoring — thresholds, trends and anomaly detection that need no failure examples — while the programme starts building the labelled record that richer models will later learn from. That phased route delivers value earlier and defers the heavier modelling until the data can support it. What thin records genuinely rule out is a supplier promising precise failure prediction from day one; treat that promise as a sign the history was never examined.
What keeps costing money after predictive maintenance models go live?+
The models themselves. Equipment is overhauled, operating regimes shift, sensors drift and get replaced — and each of those changes erodes a model trained on the plant as it used to be. Someone has to watch prediction quality, retrain on fresh history, retire models that stop earning their alerts and keep the sensor fleet healthy. Budget model operations as a standing category alongside the maintenance the programme improves; a programme without it decays into an alert stream the crews learn to ignore, which is the most expensive possible outcome of the whole investment.
In this cluster
A readiness assessment that inventories your instrumentation, data history and critical assets, so the programme you budget is the one your plant can actually support.
Walk your plant with us
