The Mantua Group

Simple Black and White Asset Management, Reliability Expertise, and Maintenance Execution Perfection.

Suspended Data: Why the Survivors Must Testify

A quick question for anyone doing failure data analysis: what do you do with the units that have not failed yet? If your answer is “nothing, they have not failed”, your reliability estimates are badly biased. Usually pessimistic, sometimes wildly so.

Unfailed units are called suspensions, or censored data, and they carry real evidence. A pump that has run 40,000 hours without failure is telling you something important about the population, even though it has no failure date to record. Classical failure-only analysis throws that evidence away, and the arithmetic of what remains systematically understates how long your equipment actually lasts. We have seen fleets where ignoring suspensions roughly halved the apparent characteristic life and triggered replacement programs that were never needed. The units still running are witnesses. Analysis that refuses to hear them convicts healthy equipment on incomplete evidence.

Knowing Your Censoring

Real maintenance datasets were rarely designed for analysis, and the gaps come in more than one flavour:

  • Right censoring. The unit is still running at the analysis date. The most common case, and the one handled by entering survivors with their accumulated hours.
  • Left truncation. Old installations whose early failures happened before records began, so the observed sample is conditioned on survival. Common in long-lived utility fleets, and dangerous when ignored.
  • Competing failure modes. A unit removed for one failure mode is a suspension with respect to every other mode. Whether to decompose by mode, and when the framework misleads, is the subject of our white paper Competing Risks, Or Not? and the companion podcast episode Competing Risks or Not.

Our white paper Missing Data, Truncation, Censoring treats the full landscape: reading reliability datasets that were never designed for analysis and drawing honest inference from imperfect records. Its companion, Dealing with Data, covers the discipline of turning reliability data into reliability decisions.

How the Tools Help You Discover the Benefit

The Weibull modules in Availability Workbench and Reliability Workbench handle suspensions natively. You enter your failures, you enter your survivors with their accumulated hours, in time-based or date-based form, and the mathematics accounts for both, by rank regression or maximum likelihood, with two- and three-parameter models available. Fitted results flow directly into system models, availability simulation and RCMCost optimisation, so the corrected estimates immediately reshape the strategy built on them. The frequent discovery: your assets are more reliable than your failure-only analysis suggested.

For readers who want the research frontier: the treatment of left truncated and right censored data in utility asset fleets, where installation records may be lost and only recent failures observed, is an active area with substantial capital consequences, and one The Mantua Group is engaged with academically. Before your organisation commits to a fleet replacement program, the analysis behind it deserves this level of scrutiny; our Weibull and survival analysis service provides exactly that second look.

Go Deeper

How can we help? Talk to us at contact@mantua.group.