Enterprise AI · Analytics

Each rung of analytics pays back only when the rung below it holds

Five rungs for Monitor, Predict, Health and Visual Inspection, with the cost of skipping each one stated beside its payback. The first three contain no analytics capability at all.

Asset management team reconciling a disputed monthly maintenance compliance figure against Maximo work records

Rung zero

Where most estates are standing when the analytics conversation starts

Before analytics helps, an organisation usually has to fix something less interesting: the fact that the same month produces several defensible numbers.

Three numbers for the same month

Operations reports one compliance figure, reliability reports another, and the figure reaching the board is a third. All three are calculated correctly. They answer three different questions that were never written down as different questions.

The argument is about definitions, not data

Does a PM completed four days late count as compliant. Does a job deferred with engineering approval count as a miss. Does a retrospectively closed work order count on the day it was done or the day it was closed. Nobody has ruled, so each team ruled locally.

Nobody can get from the number to the work

A site manager challenges the figure and asks which jobs are in it. Where that takes two days to answer, the challenge wins by default and the report loses its authority permanently.

You pay for it in meeting time spent on arithmetic rather than assets, and in the authority a report loses the first time it cannot survive a direct challenge.

The ladder

Five rungs, and what each one costs to skip

Read the cost of skipping before the requirement. Rungs 1 to 3 contain no analytics capability, which is why so many roadmaps start at rung 4 and stall. The compliance figure used to illustrate rungs 1 to 3 is illustrative.

  1. Rung 5

    Rung 5: widened on production evidence, including a decision to stop

    Requires
    Expansion argued from what the previous phase changed, in controlled phases, with the same governance in production as in pilot.
    Skip it and
    Without an explicit stop decision, a programme widens until it exhausts the attention available to review its output. Most analytics roadmaps never take that decision.
  2. Rung 4

    Rung 4: signals wired into the planning cycle

    Where a MAS analytics engagement starts paying
    Requires
    Manage data good enough to carry the question, a small number of high-consequence asset classes rather than full coverage, output aligned to planner and operations workflows, and a named reviewer on the signals that matter.
    Skip it and
    Deploy across twelve asset classes before two are trusted and you get coverage nobody acts on, which is the most common shape of a stalled analytics programme.
  3. Rung 3

    Rung 3: drillable down to the work

    Requires
    A route from the headline figure to the individual work orders inside it, fast enough to use in the meeting where the challenge is raised. On the illustrative figure, the site manager got his site’s list in the meeting, found four properly deferred and one genuine miss, and accepted the number.
    Skip it and
    A number nobody can drill into loses its first challenge and, with it, its authority. Permanently, because nobody re-litigates a discredited report.
  4. Rung 2

    Rung 2: one source, reproducible on demand

    Requires
    The figure comes from Maximo rather than a filtered extract, over a stated window, and two runs a fortnight apart reconcile. Deferrals are recorded as decisions with an approver, so a controlled deferral stops looking identical to a job nobody did.
    Skip it and
    A definition nobody can apply reproducibly is a definition in name only. The first request to run it again exposes the spreadsheet lineage and the ruling loses its force.
  5. Rung 1

    Rung 1: one definition, one owner

    Requires
    A recurring argument named, and one person with authority to rule writing the definition down. On an illustrative disputed compliance figure that was one page and four rulings: what counts as a PM in scope, what counts as on time, how an approved deferral is treated, and which date governs a retrospectively closed job. Her name on it.
    Skip it and
    Buy a capability first and it computes a number with the same confidence as the three you already have. You have added a fourth version of the disagreement.

Which rung are you on

Five statements, in order. Stop at the first one that is not true.

That is your rung. Each statement is either demonstrable this week or it is not.

  1. 1

    Somebody can name the person who owns the definition of your headline maintenance figure.

  2. 2

    The same figure, run twice a fortnight apart, reconciles without anybody having to explain a difference.

  3. 3

    A properly approved deferral is distinguishable in the data from a job that was never done.

  4. 4

    A challenge to the figure can be answered from the work orders inside it, during the meeting.

  5. 5

    At least one analytics capability you hold a licence for has been deliberately left unused, in writing, because its precondition is unmet.

If you stopped early

Most estates we meet stop at statement two or three. That is not a criticism of the people producing the numbers; it is what happens when nobody has been asked to own a definition. Start at the rung you stopped on rather than at the capability you have a licence for.

If every statement held

You are at rung 4 or above. The next decision is which asset classes to widen to and where to stop, rather than which component to buy.

Capability areas in scope

Where each IBM MAS capability joins the ladder

Four capabilities, the rung each joins, the condition that has to hold first, the decision it settles, and who is accountable for acting on its output. Licensing a capability whose precondition is unmet is the most expensive way to discover a data gap.

IBM Maximo Health: condition and criticality-led prioritisation

Usually the first capability worth attempting, and the one most exposed to data gaps. Read the Health page for the implementation detail.

Joins at
Rung 4
Precondition
Criticality and hierarchy that hold across sites
Decision it settles
Which asset gets the next intervention, and on what basis
Accountable
The reliability engineer who owns the asset class

IBM Maximo Monitor: live operating signal and anomaly context

Where OT connectivity is absent this is an instrumentation conversation rather than an analytics one. Monitor carries the procedure.

Joins at
Rung 4, where connectivity exists
Precondition
Sensor or OT connectivity to the assets being observed
Decision it settles
Whether a trend justifies an intervention before the planned window
Accountable
The named reviewer on that signal, with the planner

IBM Maximo Predict: failure probability where history supports learning

A population with thin or inconsistently recorded failures will not support it, whatever the licence says. Predict sets out the strategy it belongs to.

Joins at
Rung 5 on most estates
Precondition
Failure history consistent enough to learn from
Decision it settles
Which units in a population are prioritised, and roughly how urgently
Accountable
The reliability lead for the population

IBM Maximo Visual Inspection: image-assisted inspection workflows

Scoped to one inspection type rather than to an inspection programme. Visual Inspection works through one disagreement beat by beat.

Joins at
Independent of the ladder, scoped narrowly
Precondition
A visible defect, a repeatable capture position, labelled examples
Decision it settles
Whether an inspection finding is raised as work
Accountable
The inspector, who can override and whose override is recorded

Scope and boundaries

Decisions that stay with people at every rung

Three limits that decide whether an analytics roadmap produces decisions or dashboards.

It will not tell you what should be measured

Whether an engineering-approved deferral counts as a compliance miss is a policy question about how your organisation wants to be held to account. Analytics computes either answer with equal confidence. Choosing belongs to a named owner.

It will not overcome an unmet precondition

Predict needs consistent failure history. Monitor needs connectivity. Health needs criticality and hierarchy that hold across sites. Where the precondition is absent, the output is confident and unusable, which is worse than no output. That is data work, and it comes first.

It will not create the capacity to act

Earlier insight pays only where a planning cycle can absorb it and a maintenance resource can execute it. An estate already at the limit of its outage windows may get more from settling its reporting definitions and its backlog policy than from a new capability, and we will say so.

Advanced analytics, frequently asked questions

Where should we start with MAS analytics components?
With the recurring argument you want ended rather than with a component, because that determines both the component and the definition work that has to precede it. Most operators then start with focused Health or Monitor scope.
Can we skip a rung if we already hold the licences?
Holding the licence is not the same as holding the precondition, and the ladder is about the precondition. Where the definition is unowned or the figure is not reproducible, a licensed capability produces output that enters the same argument rather than ending it. Rung 1 is the cheapest to climb and it is not a technical exercise.
What makes a reported number trustworthy?
Rungs 1 to 3, in order: a written definition with a named owner, one source rather than several extracts with deferrals recorded as decisions, and a route from the headline figure to the records inside it fast enough to answer a challenge in the meeting.
Do you operate these capabilities in production?
Yes, through the same managed operating model used for IBM Maximo and MAS, with the MaxIron Portal as the control plane for approvals, workflow history and support evidence. Material outcomes stay gated before they affect production operation, as governance and assurance sets out.
How is this page different from the MAS component pages?
Those pages describe what each IBM capability does and how it is implemented. This page places them on a ladder: which argument each is meant to settle, which rung must hold before its output is trusted, and who owns the definition of the number it produces.

Bring the number your last three meetings argued about.

Not an analytics roadmap. One reported figure that different parts of your organisation calculate differently, and we place you on the ladder honestly: what the definition would have to say, what has to change in Maximo for the figure to be reproducible, and which MAS capability, if any, earns its place after that.

Bring this to the first call

  • One reported figure that is disputed, and the two or three versions of it that circulate
  • The name of the role that could rule on the definition and make it stick
  • How the figure is produced today, including the spreadsheets in the chain
  • The last time somebody asked which records were inside it, and how long the answer took