An asset manager running a portfolio of eighteen installations — solar, some wind, a couple of logistics hubs and an industrial plant — walks into the monthly operations committee with two figures on the dashboard: an MTBF of 720 hours, an MTTR of 12 hours. The CFO asks why last week produced three unplanned shutdowns, given that a 720-hour MTBF implies roughly one shutdown per month per asset. The asset manager has no short answer. The CMMS numbers are correct, the data is complete, the formula matches the textbook. And yet the metric does not describe the operational reality the CFO just read in the weekly availability report. The problem is not in the calculation. It is in the aggregation.
MTBF and MTTR were designed as metrics for a single asset, or for a homogeneous class of assets. When they are computed across a portfolio with mixed criticality, mixed age and mixed failure profiles, the arithmetic mean acts as a blurred lens: it returns a number that is technically correct and strategically useless. The three failures of last week may have hit three class-A assets whose real MTBF is around two hundred hours, while the average is lifted by the fifty class-C assets that have run five thousand hours without incident. The global figure is arithmetically true and diagnostically false. Across a portfolio of eighteen mixed installations, asymmetry is the rule, not the exception.
The arithmetic is right; the aggregation is not
The MTBF formula is total operating time divided by the number of failures in that time; MTTR is total repair time divided by the number of failures. Both assume that the underlying variable follows a reasonably stable distribution and that the assets belong to a population with common statistics. Neither assumption holds in a real distributed portfolio. The distribution of time between failures in a central inverter at a solar plant does not resemble the distribution in an electric motor at a logistics hub, and neither resembles that of a valve at an industrial service connection. Averaging them yields a mean weighted by asset count, not by operational impact and not by failure probability. The committee acts on impact and probability. The gap between what the metric returns and what the decision requires is what produces the monthly-committee scenes. This is not a subtle statistical point. A single mixed portfolio can show a rising monthly MTBF while critical assets are quietly degrading, because the composition of the base shifts as low-criticality assets accumulate uptime faster than they fail, and the mean drifts upward without any real change in the failure rate that matters.
What to split before you aggregate
A useful reading of MTBF and MTTR across a distributed portfolio requires splitting before aggregating. Four axes resolve most cases. The first is asset criticality, typically in A, B and C classes derived from an FMEA or RCM analysis. Criticality is not defined by the value of the asset but by the impact of its failure: a cheap component upstream of a full line is class A; an expensive component with immediate redundancy can be class C. The second is the temporal window: MTBF computed over productive hours is not comparable with MTBF computed over twenty-four hours, especially for assets with asymmetric load or strong seasonality. The third is the asset class: mixing pumps, control boards and hydraulic elements inside the same indicator dilutes the signal until it becomes useless. The fourth is age band: assets under warranty, assets in stable operation and assets in end-of-life produce different failure distributions and should not share a global metric. Splitting along these four axes turns one number into four; and those four numbers enable the operational conversation the flat average cannot sustain. A committee that starts receiving four segmented figures every month begins to ask the right questions within two cycles. A committee that keeps receiving one aggregated number keeps asking why last week did not match the average.
The data infrastructure that supports the reading
Splitting requires that maintenance data be captured with structure from the source. ISO 14224:2016, the standard for the collection and exchange of reliability and maintenance data, was born for petroleum, gas and petrochemicals, but its equipment taxonomy, its failure cause-and-consequence model and its standardized format map directly onto any industrial portfolio with mixed assets. The standard defines the minimum fields a work order must carry so that MTBF and MTTR can be recomputed by criticality, by class and by window. In operations where the current CMMS does not capture criticality or class, or captures them as free-text fields with no normalization, traceability breaks and the only reportable metric ends up being the flat average. Recovering the ability to split takes two steps: retagging the asset master with criticality and class against a coherent taxonomy, and enforcing in the field interface that a work order always opens against a typed asset. The first is a bounded project; the second is a process change.
The modular architecture of Maptainer (M01 inventory, M02 corrective, M04 field readings) allows both steps to be addressed separately and in sequence. Operations start with retagging the master without disrupting daily work; once assets are typed, the corrective module begins returning segmented KPIs with no additional change on the technician's side. Segmented reporting is not a custom development. It is the natural consequence of closing the base data with the right structure from the start.
From the average to the distribution
The useful question for an asset manager walking into the committee is not which average to report. It is which distribution to show. When operations present MTBF and MTTR split by criticality and by class, the CFO stops asking why the average disagrees with the perception of the week and starts asking where to allocate the next slice of CAPEX. That is the shift in the conversation that justifies the investment in cleaning the master and in structuring the work order. A flat average does not survive an informed operations committee; a distribution does. The asset manager who starts presenting distributions stops defending figures and starts proposing decisions, which is the role the organization hired them for.