Case studies

Patterns from 200+ Inspectaman audits.

We built Foundry Cortex on top of what failed across hundreds of factory floors. Here's what we saw — patterns anonymised, attributed to Inspectaman's audit history. Every claim traces to a real plant or a cited industry stat. No invented case-study customers.

The PdM tools on the shelf — what 73% of factories do with sensors after install

Across audits, the dominant failure mode for predictive maintenance is not the AI. It is that the sensors get installed, the dashboards get built, and within 12 months nobody opens them. The reliability champion gets promoted or retires, succession is absent, and the program decays.

Foundry Cortex's 60-day pilot exits with a documented operational rhythm — not just a working dashboard. Program ownership is structural, not personal.

Read the field report →

The Bill Problem — when senior maintenance engineers retire in Foshan

Every operations leader names a "Bill" within ten minutes of conversation. He knows which valve weeps, which compressor runs hot in the third hour of summer. He retires in 18 months. The SA engineer shortage means his replacement is not waiting in the wings.

Pilots include a Bill Inventory capture pass — 40 hours of senior-engineer reasoning indexed by asset and failure mode, so the next reliability lead inherits the pattern knowledge.

Read the field report →

Why ISO 55000 certifications don't predict reliability outcomes

Audited plants with ISO 55000 asset-management certification have failure rates indistinguishable from uncertified plants in the same sector. The certification rewards documentation discipline, not failure-mode coverage. The compliance audit is decoupled from the reliability outcome.

We measure cost-per-failure-prevented monthly against the customer's general ledger — verifiable economics, not certification theatre.

The "almost fits" problem — why operators distrust new software

The most common operator complaint across 200+ audits: "the software almost fits but not quite." A required field doesn't match the operator's mental model. A workflow assumes one shift pattern. A dashboard hides the metric the operator actually needs. Adoption stalls.

Our pilot scoping interviews start with the operator workflow — we build dashboards backwards from what the operator already does in spreadsheets, not forwards from the platform default.

Tribal knowledge loss — what disappears when a 20-year operator quits

Operators carry vendor shadow knowledge (which supplier under-delivers, which buyer can expedite), process workarounds (the 7 things the SOP says but nobody does), and relationships (the technician who answers their phone at 2am). The least-recoverable category walks out with them.

The Bill Inventory captures process workarounds and vendor relationships alongside asset history — transcripts indexed against the same historian as the sensor data.

Read the field report →

The R200k/hour failure no one budgeted for

Mid-size manufacturers carry $125k-$260k/hour downtime exposure but rarely have line-item budgets for catastrophic failures. The 2-3 events per year that account for 60% of annual downtime cost are absorbed into "operational overhead," making the ROI case for prediction harder to construct from the finance side.

Our pilot proposal includes a downtime-exposure model the CFO can verify against the general ledger — failure cost surfaces as a real line item, not absorbed overhead.

Tier-2 supplier hidden costs — what IATF audits miss

IATF 16949 audits focus on quality systems documentation. They do not audit reliability spending or downtime cost. Tier-2 automotive suppliers carry hidden reliability debt that doesn't surface until a Tier-1 quality incident triggers a re-audit, by which point the cost is order-of-magnitude larger.

For automotive tier-2 customers, our quarterly review surfaces reliability debt against IATF audit cycles — visible before the Tier-1 quality incident.

Read the field report →

Load-shedding's invisible damage to bearing life

SA plants running on intermittent grid power see motor starts 3-5× more often than steady-state plants. Each start carries bearing-load and lubrication stress invisible to monthly lab analysis. The cumulative damage shortens bearing life by 30-40% — but doesn't show up until catastrophic failure.

Our motor-current signature models track start-cycle counts per asset; the load-shedding stress signature surfaces before bearing failure, not after.

By sector — what we see most.

Mining

  • ·Conveyor bearing failures dominate — accounts for ~40% of preventable downtime
  • ·Load-shedding-induced motor stress is invisible to lab analysis but visible in current signatures
  • ·Tribal-knowledge concentration — 60% of plants have one senior reliability engineer with no successor

Manufacturing

  • ·IATF 16949 compliance forces documentation that masks real failure patterns
  • ·Press dies wear differently per product mix — generic models miss it; per-asset models catch it
  • ·CMMS adoption is widespread but work orders rarely link to predicted-failure forecasts

Food & Cold Chain

  • ·Hygienic sensor housings add 15% sensor capex — recoverable in 6-8 prevented failures
  • ·Compressor failures during peak summer are the highest-cost incidents (R200k+ per event)
  • ·Cleaning-cycle SIP/CIP corrosion patterns are detectable 4-6 weeks ahead in vibration trends

Energy & IPP

  • ·Inverter fan failures cause cascading thermal events; vibration + temp catches them 5-10 days out
  • ·Solar tracker drive failures track sun angle — predictable seasonal pattern, missed by reactive maintenance
  • ·Grid-tie protection relays drift over 3-5 years; routine re-cal misses the early signal
Next step

Put a robot on your site in 90 days.
Fixed price. Fixed scope. No "let's get on a discovery call" loop.

A supervised pilot with our operators and your team — feasibility, deployment, and a board-ready evidence report. Honest scope, honest price, honest verdict. Tell us your site and we come back within two working days with a scoped proposal — what we'd deploy, the tasks we'd run, the safety case, and the evidence we'd target.