£2.4m
£5.8m
Lost every year
To failures your alarms
never saw coming.
Where the £2.4m – £5.8m comes from.
Unplanned line stoppages and batch loss.
Thermal efficiency drift on boilers and process heat. Silent failure.
Callouts, overtime, expedited parts, agency cover.
Smaller sites scale down proportionally (£400k – £1.2m for 50 – 100 staff). Larger sites scale up (£5m+ for 500+ staff).
Does this apply to your site? Yes.
We catch departure from healthy before any alarm trips — using data your historian already stores. No installation. No disruption. Four weeks.
If you already have PI (OSIsoft), FactoryTalk Historian, AVEVA Wonderware, or a predictive-maintenance vendor, you're ahead of most of the UK manufacturing base. Those tools monitor thresholds and symptoms — not departure from healthy. The gap isn't solved; it's just smaller.
How much of this is actually recoverable.
Documented recovery ranges from UK and global manufacturing — sourced ranges, not vendor projections.
Conservative end of McKinsey's 30 – 50% range, realistic for a first deployment at a mid-sized UK site.
Recoverable thermal efficiency loss from fouling, scaling, and combustion drift.
The multiplier on labour rates when faults are caught at escalation rather than early-warning stage.
UK wholesale gas price, late February → late March 2026.
Hormuz disruption. Still volatile.
A 5% thermal efficiency drop never trips an alarm.
It's within spec. No flag, no callout, nothing to investigate. But at today's UK gas prices, on a 10 MW site, it silently costs £300,000 a year.
The one your current monitoring cannot see.
The method is already proven.
The approach has been validated across five public benchmark datasets.
No fault labels required. No run-to-failure training data. No retraining between similar assets.
This page covers the outcome claims and the public benchmarks they were validated on. The underlying methodology is proprietary and not published.
Boiler fouling caught without fault labels.
Three severity levels. No labelled training data. All three correctly ranked from the commissioning baseline — the mild case included, which conventional alarms do not flag.
LBNL Boiler Plant public benchmark. Full method summary available on request.
95% = first indication, no alarm triggered. 80% = clear departure. 65% = intervention required. Every level detected from a window of healthy operation alone.
Seven days. One asset. What your current monitoring missed.
Six to twelve months of one asset's historian export. Whatever sample rate you already log at. CSV, archive, database dump — doesn't matter.
We identify a window of healthy operation in your own record, calibrate the classifier against it, and run it across the rest looking for departure events.
A concise written note: what we detected, when, what it likely is, and whether your maintenance records show the event downstream. Working shown, caveats stated.
What we do with your data.
Click to expand.
+
Transferred via SFTP under standard mutual NDA. We send you the endpoint credentials; nothing is pushed from your side without a direct engineering contact.
Stored on UK-based encrypted infrastructure for the duration of the evaluation only.
Not used for model training, benchmarking against other sites, or any secondary purpose. Not shared with any third party.
Deleted within 14 days of the findings note being delivered, unless you ask us to retain it for a follow-on engagement.
A named contact — typically a plant manager or engineering lead who can answer a handful of process questions if our findings need context.
An honest 20-minute call before we start: what the asset is, what you already know about it, and what a surprise finding would look like.
If the findings are useful: we can scope a full pilot across multiple assets. That's a separate conversation and a separate engagement.
If they aren't: you keep the findings note, we delete the data, and we part on good terms.
The £2.4m – £5.8m headline is a composite estimate built from seven cited sources.
Every number's source. Click to expand.
+
46% of UK manufacturers report 6 – 10 unplanned incidents per week; 15% report 11 – 20.
Fluke Corporation / Censuswide, The True Cost of Downtime — UK Manufacturing, October 2025. Survey of 600+ senior decision-makers across UK, US, Germany.
45% of outages last up to 12 hours; 17% stretch to 72 hours. This is the frequency floor the breakdown figures are built on.
£2,000 – £8,000 per hour of single-line stoppage at a 300 – 500 employee batch plant.
Derived from ONS Annual Business Survey (manufacturing gross value added per employee, 2023 – 24) and ATA Recruitment salary benchmarking (2024 – 25). Wider sector averages from Siemens True Cost of Downtime 2024.
Lost margin + idle direct labour + materials/WIP scrap + energy continued + expediting. Higher for speciality chemicals or pharma-adjacent food; lower for simpler lines.
5 – 15% cumulative efficiency loss from fouling, scaling, and combustion drift when boiler maintenance is deferred.
Carbon Trust industrial heating guidance (archived BEIS/DESNZ evidence base); consistent with ASHRAE boiler efficiency guidelines and DESNZ industrial heat decarbonisation evidence.
Most sites do not measure this because it is drift, not failure. Thermal KPIs stay within spec while efficiency erodes silently.
3.5 – 4.4p/kWh baseline (Q3 2025). Market now materially higher and volatile.
DESNZ Quarterly Energy Prices, September & December 2025 (manufacturing industry band). Post-conflict context: House of Commons Library Economic Update: Middle East Conflict, April 2026 — UK wholesale gas rose ~75% between late February and late March 2026 following Hormuz disruption.
A 1p/kWh shift moves the thermal-recovery component by roughly 20%. At Q1 – Q2 2026 rates, a drift loss that was a £300k problem becomes £500k without the plant changing anything.
20 – 30% downtime reduction from predictive maintenance programmes (conservative end of McKinsey's 30 – 50% range).
McKinsey & Company, Manufacturing: Analytics unleashes productivity and profitability, August 2017. Reinforced by Deloitte and ABB case analyses 2022 – 2024.
The 30 – 50% headline applies to mature implementations at enterprise scale. For a first deployment at a mid-sized UK site, 20 – 30% is the realistic expectation.
Only 12% of UK manufacturers currently use predictive maintenance.
Fluke Corporation / Censuswide, The True Cost of Downtime — UK Manufacturing, October 2025.
88% of the UK manufacturing base has not yet implemented these approaches. This pain is active, not hypothetical.
UK multi-skilled shift maintenance engineer base pay £44k – £52k/year.
ATA Recruitment, Annual Engineering Insights Report 2024 (published March 2025). Live vacancy data from ATA Recruitment and Hays Manufacturing Engineering, 2025.
OTE including shift allowances and overtime reaches £52k – £60k+ for experienced staff. Emergency callouts typically run 2 – 3× standard rate. Agency cover during vacancies adds 15 – 25% on top of salary.
Energy ranges above use the Q3 2025 DESNZ industrial baseline. Current Q1 – Q2 2026 market rates sit materially above that baseline after the Hormuz supply disruption. The high-end figures on this page understate exposure at today's gas prices.
We're selecting three to five sites for our first reference pilots.
A written findings note on one of your assets — what we detected, when, and whether your records show the event downstream.
Seven days. One asset. The honest answer to whether early-warning detection surfaces something your current monitoring is missing on your specific site.
No cost.
A reference case — with your agreement, anonymised as required.
Our first reference pilots matter to us. In exchange for being one of them, we do the work at no cost to you and we put real effort into making the findings hold up under scrutiny.
Three to five sites, 2026. Because this first set has to be right.
Seven days. One asset. Your findings.
Standard mutual NDA. UK-based encrypted storage. No installation on your site. Data deleted within 14 days of delivery. Full handling detail in the section above.