A 4-week read-only shadow pilot on your historian and alarm exports. We agree the KPI model first. If the sealed evidence pack maps $5M+ in addressable operational value, the pilot fee applies — $35K to $75K per process unit, priced by data volume, security requirements, and pilot success criteria. If not, you keep the pack and pay nothing.
A 4-week read-only shadow pilot on historian and alarm exports. We agree the KPI model first. If the sealed evidence pack maps $5M+ in addressable value, the pilot fee applies. If not, you keep the pack and pay nothing.
Real questions your engineers ask when the console lights up. Twin-3 was engineered against these four conversations. Each one costs measured money. Each one gets an audited answer.
Between a vendor guess and a verified answer sits an audited evidence layer.
"Which actuator surprises me this week."
Twin-3 reads the valve, controller, and upstream signature in your historian. The bearing drift is in the data your plant already collected — Twin-3 surfaces it as ranked, math-verified evidence, so your team acts before the unit derates.
"Loop is cycling. Valve, controller, or upstream — the historian has the answer."
Twin-3 traces the cycle to the tag that started it, with ranked attribution sealed in the evidence pack. The historian stops being a haystack. Your engineers stop hunting. The third day of investigation stops costing more than the second.
"The auditor will ask why we missed it. I need evidence, not a screenshot."
Twin-3 returns a SHA-256 sealed evidence pack. Procurement IT runs the hash check in ten seconds. The board sees what was seen, when, and why. A dashboard screenshot isn't procurement evidence — a sealed pack with full chain-of-custody is.
"70% of our alarms are nuisance. The console stopped being trusted years ago."
Twin-3 ranks alarm floods into a worklist, not noise. Operators see the signal first. The console becomes trusted again. The sealed pilot pack maps each alarm cluster to recoverable engineering hours — your team gets the rationalisation programme they've been writing for years, in four weeks, on your own data.
Every Twin-3 finding travels two phases. The Detective searches your historian for the law that broke. The Vault seals what was found so procurement, the regulator, and the board can verify it themselves. The arc reduces the time between alarm and answer.
The detective beam scans your refinery's digital nervous system, ignores the nuisance traffic, and snaps onto the specific bearing, valve, or controller that is about to break the loop. Every step traces back to a physics equation the plant already obeys.
The diagnostic compresses into a sealed evidence pack. A SHA-256 hash crawls across the surface and locks. Timestamps, sensor logs, ISA-18.2 / IEC 62682 alignment checks. Filed to a read-only archive. Audit-ready. Procurement-verified.
Twin-3 does not guess your plant. It reads the laws your plant already obeys, and tells you, with proof, the moment those laws break. Every residual sits inside a conformal coverage bound. Every coverage bound sits inside an audit chain. Every audit chain sits inside a hash you can verify.
No black box. Every threshold and coverage bound is disclosed in the evidence pack — your procurement team reviews them before any deployment conversation.
Engineering direction 11 of 12 fault families today, with IDV3 isolated as a known boundary case. The engineering work continues to push that boundary toward as-close-to-entirely as the physics will allow. The product is benchmarked, sealed, and ready for bounded pilot evaluation. The next benchmark release tightens the boundary further.
Twin-3 was sealed on two structurally different public fault-detection benchmarks (Tennessee Eastman + DAMADICS), plus a third public dataset (Manca) used to structure alarm-burden behaviour. Every number below sits inside a SHA-256 evidence chain procurement IT can verify in ten seconds. Real-plant performance is established during your pilot, on your own historian — that’s the point of the pilot.
No. Twin-3 is read-only on historian, alarm, and sensor exports. No DCS access. No SIS access. No control authority. No replacement of certified safety systems.
The deployment surface is your data export, not your control network. No OT asset modification. No DCS install. No control-network access. Your existing IT, security, and procurement workflows still apply — DPIA, vendor onboarding, data-handling review — but the pilot does not touch any safety-rated or control-rated infrastructure.
Three CSV exports: tag list, historian sample, alarm journal. 4-week window. Hosted in your jurisdiction of choice. Deletable on request at any point including after pilot.
No SCADA tunnel, no VPN into your control room, no agent installed on a plant machine. The pilot is procurement-grade because the data surface is procurement-grade — three files, one analysis environment, full chain-of-custody record.
Every artefact in the pack carries a SHA-256 hash and a manifest. Your IT team verifies the chain in 10 seconds with standard tools. No special verifier required.
The pack is designed for procurement IT before it is designed for engineering. Audit trail, timestamp, hash chain, derivation log, ISA-18.2 / IEC 62682 alignment markers, GDPR-compatible. If it fails a review, the deployment never starts — and we want to know.
Those platforms run inside your plant infrastructure — often integrated with your historian and DCS architecture. Twin-3 runs read-only on a static historian export, outside that infrastructure. Different scope, different deployment, different procurement path.
We do not replace your APM or your historian. We sit beside them and answer one question: where in your existing data is the addressable operational value you have not yet acted on. The output is forensic evidence, not control output.
If the sealed evidence pack does not map at least $5M in operator-agreed addressable value, the pilot fee does not apply. You keep the evidence pack. There are three commercial outcomes, defined before any data is sent:
Above $5M. Pilot fee applies, you keep the sealed evidence pack.
Below $5M, no credible path. You pay nothing and keep the evidence pack.
Below $5M, credible path found. You pay nothing for the first pass. Twin-3 may fund one agreed diagnostic expansion under a written extension. If the second pass still does not clear the value gate, no fee applies and the engagement closes.
We are not guaranteeing realised savings before implementation — realising value depends on your team's maintenance and operations decisions. We are guaranteeing a value-gated pilot: before it starts we agree the KPIs and cost model; Twin-3 then runs read-only and produces sealed evidence against that model. Twin-3 does not ask you to believe a claim — it asks you to test the claim on your own data, under your own KPIs, with a payment gate tied to verified addressable value.
Public-benchmark coverage today: continuous chemical process units (Tennessee Eastman) and pneumatic actuator-fault detection (DAMADICS official events). Equipment beyond these two benchmarks is scoped explicitly in the pilot — declared coverage, not assumed coverage.
We do not over-promise on equipment families we have not benchmarked. If your unit type sits outside the public-benchmark coverage, the pilot starts with a scoping conversation, and you keep the evidence pack either way. The current sealed build covers 11 of 12 Tennessee Eastman fault families and 19 of 19 official DAMADICS actuator-fault events.
Contract to evidence pack: 4 weeks. Eight hours of your team's time. Week 1 scoping, weeks 2-3 forensic run, week 4 evidence pack delivery and value review.
No multi-month integration project. The deployment is bounded because the data surface is bounded — three CSV files, one analysis environment. If we slip the 4-week window without your prior agreement, the pilot becomes free regardless of evidence.
You. Always. Read-only means we do not retain. The evidence pack is your property under the pilot agreement. We retain only the methodology (the sealed-build release artefacts) and the public-benchmark calibration.
Your operational data is yours. The derived findings are yours. The hash chain proves chain-of-custody. The methodology and benchmark calibration are ours — that is what we license, that is what the fee is for. Nothing about your plant lives in our systems after the pilot closes if you so request.
The standard Twin-3 pilot pack is structured under English law with UK data-protection controls. Operator-local data requirements are handled through a jurisdiction schedule before any export is received.
For UAE operators, a UAE PDPL jurisdiction schedule is drafted to cover cross-border transfer and sharing requirements before any export is received. For Saudi operators, a KSA PDPL schedule is drafted to cover controller obligations and rights. Dispute routes can be set to DIFC or ADGM arbitration where operator procurement requires it. Data residency is operator-selected: UK secure cloud, operator cloud tenant, or local/GCC-hosted environment, defined before data transfer.
Four fields. A defensible range, not a vendor estimate. Every multiplier sourced. Every assumption visible. The calculator below answers "would Twin-3 cross the $5M threshold on our unit?" — auditable in 30 seconds. Forward the URL to your reliability lead; the configuration is preserved.
No pilot fee if we cannot map the value.
Paid only when the evidence clears the gate.
4 weeks of historian data. 8 hours of your team’s time. You receive a sealed evidence pack mapped against KPIs you already use. Pilot fee applies only when the evidence clears $5M+ in operator-agreed addressable value. If it doesn’t, you keep the pack and pay nothing.
Unit FC-247 · Alarm Journal · 28-day window. Twin-3 identified a chattering cluster across TI-301 / PIC-204 / LIC-208 generating 73% of console noise during operator shift handovers. Pattern matches priority-inversion signature documented in ISA-18.2 §13. Hash-chain verified. Source rows attached.
Rationalisation of 11 alarm setpoints plus control-loop tuning on PIC-204 and deadband adjustment on TI-301. Estimated implementation effort: 2 engineering days. Estimated time-to-impact: 1 production shift after tuning applied.
$2.3M / year addressable value, derived from: 14,200 alarm events × 4 min loaded operator-handling time × $185/hr loaded labour (your figure), plus avoided event-response cost (your figure). Pre-realisation. Booked only if action is taken.
Everything above resolves into one artefact. The pack groups every finding into five value buckets — each traced to a verifiable source, every dollar linked to an auditable formula. Here is what is inside before you commit to anything.
Unit FC-247 · Alarm Journal · 28-day window. Twin-3 identified a chattering cluster across TI-301 / PIC-204 / LIC-208 generating 73% of console noise during operator shift handovers. Pattern matches priority-inversion signature documented in ISA-18.2 §13. Hash-chain verified. Source rows attached.
Rationalisation of 11 alarm setpoints plus control-loop tuning on PIC-204 and deadband adjustment on TI-301. Estimated implementation effort: 2 engineering days. Estimated time-to-impact: 1 production shift after tuning applied.
$2.3M / year addressable value, derived from: 14,200 alarm events × 4 min loaded operator-handling time × $185/hr loaded labour (your figure), plus avoided event-response cost (your figure). Pre-realisation. Booked only if action is taken.
A successful pilot is the start of a working relationship, not the end of an engagement. The same discipline that runs week one runs year three. The operator pays only when each cycle’s pack maps agreed value — nothing else changes.
The first pilot proves the method on one process unit. The second runs on a paired unit, often within the same shutdown cycle. By month nine, the pattern is repeatable across the plant — same export protocol, same SHA-256 manifest discipline, same read-only boundary. Each new unit shortens the next.
After the first pack, Twin-3 fits into the operator’s existing rhythm — quarterly reliability reviews, pre-turnaround diagnostics, post-incident replays. Each cycle produces a fresh sealed pack against the same operator-agreed value model. Year-on-year, the operator builds a verifiable record of what was found, what was acted on, and what value was realised.
Each cycle expands the operator-confirmed pattern library for that plant. Recurring fault signatures get caught faster on cycle three than cycle one — not because the algorithm changed, but because the operator-validated reference set grew. The discipline does not change; the resolution does.
Twin-3 starts as a low-friction evidence service. It stays one — by design. There is no platform commitment, no licence ladder, no enterprise sale at year three. The relationship compounds because the discipline compounds.
Start with one unit. See whether the evidence pack clears your gate. Then decide whether the second unit is worth running.
Book a scoping callA quieter console. A worklist your operators trust. A sealed pack your auditor accepts. Engineering meetings that move forward instead of looping.
The historian becomes an addressable map. Every finding traces back to a law your plant already obeys. Your board sees the numbers, your regulator sees the evidence, and your engineering team gets back the one resource the industry never gave them — time.
Request the Scoping Briefing