Contact us

Twiniti Wiki / Enterprise Asset Management / Failure modes and ROI traps

Failure modes and ROI traps

EAM / CMMS programmes fail as change programmes, not as missing modules. Practitioner and vendor write-ups converge on the causes. No study measures a 70% CMMS failure rate. Vendor and consultant pages recycle a round organisational-change percentage onto CMMS. The named-sample CMMS studies that do exist measure dissatisfaction or under-use, not failure; they are in §8.1. The organisational-change trace is also in §8.1 so that percentage is not treated as a measured CMMS result.123

A defensible value case is a small set of business metrics over three to five years, with the programme phased so master data and actuals exist before predictive models. Return-on-investment (ROI) traps are listed as one-liners in §8.5.45

8.1 The recycled failure-rate figure — organisational change, not a CMMS sample

Vendor and consultant pages apply a round “70% fail” (sometimes “60–80%”) to CMMS implementations without naming a CMMS sample:

What named samples actually measured. Jones (1994), reviewing 725 maintenance-management audits run with one consultancy’s audit method, reported that over 60% of companies were not satisfied with their computer-aided maintenance management system and could not demonstrate clear benefits.18 A self-selected web survey by the Plant Maintenance Resource Center (87 valid responses, 2000) found 20% rated their implementation “poor” and 20–40% could not identify any benefit.19 Reliable Plant (2006) reported Kris Bagadia’s survey of 299 maintenance professionals: 94.7% said they were not using their CMMS to its maximum capability; Bagadia’s “80 percent of installations fail” is his own estimate, not a survey result.20 These are dissatisfaction and under-use measures from the 1990s–2000s, not audited outcome studies, and none supports a “70% fail” figure.

Trace. ReliaMag’s digital-transformation failure-rate guide, reading McKinsey and Hughes together:213

Wiki rule. Say EAM value is lost when adoption, data and process are weak. Do not treat a round organisational-change percentage as a measured CMMS implementation statistic unless a named CMMS sample is cited. Reliable Plant / vendor pages in this pack are not that sample. Quoting the round number as if it were a measured CMMS constant is itself an ROI trap (§8.5 item 10).

8.2 Causes that are consistently described

Independent enough overlap across TRM’s IBM-panel write-up (Goetz, 18 March 2026, vendor), eWorkOrders’ “8 ROI pitfalls” (vendor), SAMEX (vendor), UpKeep (vendor), and Maximo-Users.net’s 2026 practice notes (practitioner community, Maximo-specific) to list causes, not a rate. Every row is a process failure. None is a missing SKU.458610

Failure modeWhat it looks likeWhy ROI diesWho describes it
IT rollout, not work-process changeSuccess criterion is “go live.” Today’s chaos, faster.TRM’s first question: “Are we implementing software, or improving how we run the plant?”TRM4
No executive owner of outcomesMaintenance manager’s project; resets at every leadership changeNo one owns the three-to-five-year value case.TRM; eWorkOrders pitfall 545
Lift-and-shift dirty master dataDuplicate assets, “MISC” item master, PMs on retired kit, no bills of materialsUsers abandon the system. Maximo-Users.net names dirty data as the most consistent Maximo-failure cause they record.TRM Q3; eWorkOrders pitfall 2; Maximo-Users.net4510
Actuals / failure-code habit never formsFirst 90 days of free text and skipped actualsCost reports, job-plan accuracy and mean time between failures (MTBF) are all fiction. Hard to retrofit.Maximo-Users.net (90-day window); ISO 14224 logic on mode ≠ mechanism ≠ cause1011
Big-bang every moduleInventory + predictive maintenance (PdM) + mobile + workflow on day oneNothing is used well; the programme is “the system.”SAMEX 2.3; eWorkOrders: start with work order + PM85
Training as a one-hour click-tourNot role-scenario (“close this pump change-out”)eWorkOrders calls adoption the single biggest ROI factor — vendor FAQ, cause is still consistently named. Superusers are the usual mitigation.eWorkOrders5
Parallel systems left aliveWhiteboard + WhatsApp + spreadsheet; the EAM is an afterthoughtData stays junk. The EAM has to be the only path for authorised work.Maximo-Users.net10
Automating a bad processInconsistent priorities, no job plans, tribal storeroomDigitising chaos.SAMEX 2.68
Desktop-only for a mobile craftClose-out at the shop PC at shift endActuals and failure mode are reconstructed from memory. Mean time to repair (MTTR) clocks are lost if in-progress is skipped.eWorkOrders pitfall 7; SAMEX 2.5; UpKeep586
No post-go-live measurementProject “done” at cutoverNo 3–5 year value case; no ops-review metrics.eWorkOrders pitfall 8; TRM54
TCO surpriseLicence-only budget; migration, training, integrations extraThe business case was never the full cost.eWorkOrders pitfall 65
Wrong-fit productToo heavy (unused modules) or too light (re-migrate in three years)Chapter 7’s diligence questions exist because this is common.eWorkOrders pitfall 45
PdM / AI on garbageSensors and models on unclassified assets, no failure codes, no populationAPM cannot deliver. ISO 14224: population is the denominator; omitting churn inflates MTBF.TRM; ISO 14224 catalogue-level logic411
Planner used as clerk / firefight spareNo work packages, no ready backlogPlanning never gets ahead of the day. Doc Palmer’s planning/scheduling split (handbook; cited via AccendoReliability and ReliaMag) is the named counter-practice — modelled, not a multi-plant randomised trial.Palmer via ReliaMag / AccendoReliability1213
Reliability and maintenance fusedRoot-cause analysis (RCA) is “next week” foreverThe pager always wins. Fabrico (vendor, labelled) is conceptually clean on the split.Fabrico14

Those rows are descriptions, not a survey. They do not add up to a percentage.

8.3 A vendor-reported plant anecdote (labelled; not a Twiniti result)

TRM’s 18 March 2026 write-up of an executive panel (named panelists: Dick DeFazio, Bob DiStefano, Tom Wogenrich of IBM Maximo Center of Excellence, Ray Miciek) reports a chemical-plant anecdote: a site at 55% of nameplate with about $8 million/year maintenance, after process change plus EAM configuration, reached 110% capacity with maintenance under $3 million in three years.4

That anecdote is vendor-reported and was not independently audited for this pack. It is not a Twiniti testimonial, not a Maximo partner claim, and not a plant outcome this wiki can own. If later editors keep it at all, attribute it to TRM’s panel write-up and keep the numbers inside that attribution. Do not generalise it into an industry ROI constant.

eWorkOrders’ FAQ claim that “many companies see positive ROI in 12–18 months” is vendor, no sample. It is not used here.5

8.4 A defensible value case

TRM’s listed levers (reasonable, still vendor): maintenance spend, unplanned outages, inventory turns, asset life, warranty recovery — over three to five years, a small set of business metrics, not an ROI slide at cutover.4

ReliaMag / Society for Maintenance & Reliability Professionals (SMRP) names that belong on an ops review, with the cautions already in chapter 5: availability, MTBF (trend, repairable items), maintenance cost as a percentage of replacement asset value (RAV) — target lives in SMRP, do not invent 1–3% — planned-work percentage, PM compliance with a stated on-time rule (SMRP 5.4.14 is count, not hours; best-in-class “above 90%” is a prescribed target, not a measured industry average), schedule compliance (and Palmer’s caveat that chasing ~90% on a fully loaded week may mean the week was under-loaded), MRO value as % of RAV (ReliaMag lists the metric and refuses to reprint the paid target).15161712

Gaming remains in scope. Auto-close overdue PMs, back-date completions, or loosen the on-time window and “compliance” rises while MTBF falls. Pair PM compliance with Yield (corrective hours found via PM/PdM ÷ PM/PdM hours; SMRP: no universal target) and a reliability outcome.1615

Phasing that matches the sources:

  1. Hierarchy + item master + work-order actuals + PM generation.
  2. Job plans, failure codes, kitting, weekly schedule ritual.
  3. PM right-size from failure data (RCM / FMEA decides what belongs on the PM; the EAM cron only generates it — chapter 5).
  4. Condition-based maintenance (CBM) on a small critical set.
  5. Predictive only after population + mode data exist (ISO 14224 catalogue-level: population is the denominator).11104

Maximo-Users.net’s maturity model (community, not IBM): L1 logbook → L2 planned work → L3 failure-code-driven reliability → L4 CBM → L5 predictive / prescriptive. They locate most Maximo 7.6 sites without a reliability programme at L2. Advancing L1→L2 is process, not licences.10

8.5 ROI traps (one line each)

From the ops briefing §7.4, restated so each trap points at a cause in §8.2 or a figure this chapter refuses:

  1. Buying features instead of defining the win condition.
  2. Migrating dirt.
  3. Skipping role-based training and superusers.
  4. Leaving shadow systems on.
  5. Counting PM compliance without Yield and MTBF.
  6. Stocking by habit (ABC-only, quoted lead times, no BOM).
  7. Calling go-live “value.”
  8. Staffing planners as overtime clerks.
  9. Turning on AI / PdM before ISO-14224-quality failure data.
  10. Quoting 70% failure / 30–40% planning-time reduction / 12–18 month ROI as if they were measured constants.

Item 10 is the editorial rule in one line. Maximo-Users.net’s claim that organisations using job plans report 30–40% less planning time for recurring work is a community claim with no method shown and is not quoted as a measured industry result (chapter 5 already flagged it). eWorkOrders’ 12–18 month ROI is skipped (§8.3). The 70% figure is traced, not reprinted as CMMS fact (§8.1).10513

Sources

  1. McKinsey, Ewenstein, Smith, Sologar, “Changing change management,” 1 Jul 2015 — https://www.mckinsey.com/featured-insights/leadership/changing-change-management — 70% sentence is about organisational change; no method, sample or footnote on that sentence.
  2. ReliaMag, “Do 70% of Digital Transformations Fail? Tracing the Number” — https://reliamag.com/guides/digital-transformation-failure-rate/ — trace of the round number; distinguishes McKinsey 2015 (change) from McKinsey 2018 (digital survey). Fetched/used 28 Aug 2026 pack.
  3. Hughes, M. (2011). “Do 70 Per Cent of All Organizational Change Initiatives Really Fail?” Journal of Change Management 11(4), 451–464. DOI https://doi.org/10.1080/14697017.2011.630506 — publisher page: https://www.tandfonline.com/doi/abs/10.1080/14697017.2011.630506 — University of Brighton record: https://research.brighton.ac.uk/en/publications/do-70-per-cent-of-all-organizational-change-initiatives-really-fa/ — no valid and reliable empirical evidence for the 70% organisational-change narrative.
  4. TRM / IBM panel write-up, Will Goetz, 18 Mar 2026 — https://trmgroup.com/resource/why_eam_implementations_fail_how_to_fix_them/ — vendor. Causes; 3–5 year value case; chemical-plant anecdote is vendor-reported, not independently audited, not a Twiniti testimonial.
  5. eWorkOrders, “8 CMMS ROI pitfalls” — https://eworkorders.com/cmms-roi-pitfalls/ — vendor. Used for named pitfalls (executive owner, dirty data, big-bang, training, TCO, wrong-fit, desktop-only, no post-go-live measurement). 12–18 month ROI FAQ skipped (vendor, no sample).
  6. UpKeep, “Most common failures in CMMS implementation” — https://upkeep.com/learning/most-common-failures-in-cmms-implementation/ — vendor. Recycles 70% with no named CMMS sample; still usable for the desktop-close-out / adoption cause.
  7. Fabrico, “Why 70% of CMMS Implementations Fail (And How to Succeed)” — https://www.fabrico.io/blog/cmms-implementation-failure-reasons/ — vendor. Recycles “analysts estimate that between 60% and 80%” with no named CMMS sample. The Fabrico page used for the role split is 14, a different URL.
  8. SAMEX, CMMS implementation failure causes — https://www.samexsys.com/kc-en/cmms-implementation-failure-causes-en/ — vendor. Big-bang, digitising chaos, desktop-only; analogises to McKinsey 70% of transformations and a Reliable Plant ~80% claim — not a named CMMS sample for this wiki.
  9. McKinsey, “Unlocking success in digital transformations,” 29 Oct 2018 — https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/unlocking-success-in-digital-transformations — Global Survey: 16% improved and sustained; 7% improved but did not sustain. Not a CMMS sample. Self-reported.
  10. Maximo-Users.net, work-order practices 2026 — https://www.maximo-users.net/blog/maximo-work-order-best-practices-2026/ — practitioner community, Maximo-specific. Dirty-data as failure cause; 90-day actuals window; shadow systems; maturity model L1–L5 (also https://www.maximo-users.net/cmms-best-practices-enterprise/). Job-plan “30–40% less planning time” is community, no method — not used as a measured result.
  11. ISO 14224:2016 catalogue — https://www.iso.org/standard/64076.html — Petroleum, petrochemical and natural gas industries — Collection and exchange of reliability and maintenance data for equipment. Population as denominator; failure mode ≠ mechanism ≠ cause. Catalogue-level only; ISO.org HTML Cloudflare-blocked 28 Aug 2026. Consultancy navigator (labelled, not the standard): iFluids, 20 Jun 2026 — https://ifluids.com/standard/iso-14224-reliability-failure-data-guide/
  12. ReliaMag, maintenance staffing ratios — https://reliamag.com/guides/maintenance-staffing-ratios/ — Palmer planner ratios and wrench-time model. Not a multi-plant randomised controlled trial. AccendoReliability, Sondalini review of Palmer 3rd ed. — https://accendoreliability.com/review-maintenance-planning-scheduling-handbook-doc-palmer/
  13. Palmer, R.D., Maintenance Planning and Scheduling Handbook, McGraw-Hill — cited via 12, not quoted from the paid book.
  14. Fabrico, maintenance engineer vs reliability engineer — https://www.fabrico.io/blog/maintenance-engineer-vs-reliability-engineer/ — vendor, conceptually clean on fusing the pager with RCA. Not a failure-rate study.
  15. ReliaMag, maintenance/reliability KPI reference — https://reliamag.com/guides/maintenance-reliability-kpi-reference/ — maps metrics to SMRP / EN 15341; refuses unsourced “world-class” numbers.
  16. ReliaMag, PM compliance rate benchmarks — https://reliamag.com/guides/pm-compliance-rate-benchmarks/ — SMRP 5.4.14 best-in-class “above 90%” verified as a prescribed target, not an industry average; Palmer vs SMRP on schedule compliance; gaming.
  17. SMRP Best Practices — https://smrp.org/learning-resources/smrp-library/best-practices-metrics-guidelines/ — 7th ed. member download observed 28 Aug 2026. Do not paste unpublished numeric targets from memory. Cost-as-%-RAV target lives here.
  18. Jones, R. H. (1994). “Computer-aided maintenance management systems.” Computing & Control Engineering Journal 5(4), 189–192. DOI https://doi.org/10.1049/cce:19940405 — 725 AMIS audits; over 60% not satisfied / no demonstrable benefits. Abstract via OpenAlex/Crossref; not a “failure” rate.
  19. Plant Maintenance Resource Center, “CMMS Implementation Survey Results – 2000” — http://www.plant-maintenance.assetivity.com.au/articles/CMMS_survey_2000.shtml — 87 valid responses; 20% rated implementation poor; 20–40% no identifiable benefit. Fetched 5 Oct 2026.
  20. Arnold, P. V. (Noria), “Few make the most of their CMMS,” Reliable Plant, 31 May 2006 — https://www.reliableplant.com/Read/1627/cmms — Bagadia survey n=299; 94.7% under-use; 80% “fail” is Bagadia’s estimate. Fetched 5 Oct 2026.

Explicitly out of this chapter. “70% of CMMS implementations fail” as a measured CMMS statistic; eWorkOrders 12–18 month ROI; Fabrico/Cryotos “RCM cuts 30–40% of PMs”; Maximo-Users.net 30–40% planning-time reduction as an industry result; TRM plant numbers as a Twiniti or IBM-partner outcome; remaining seats; partner directories.