Platform
One platform from sensor reading to executive decision
Trust Quality Assurance sits above the systems you already run. It acquires quality signals at the edge, binds them to a governed model, evaluates them with statistics and machine learning, and turns the result into owned, auditable action.
- production lines connected
- 180+
- signal to alert latency
- < 2 s
- records processed per year
- 2.4 B
- enterprise availability SLA
- 99.95%
Reference architecture
Four layers, deployed where each one belongs
Latency-critical work stays in the plant. Cross-plant analytics runs in the cloud. Decision authority stays with your quality organisation. Select a layer to inspect what runs there.
Each layer can be deployed independently. Acquisition always stays inside the plant network.
Reads from the equipment, gauges and inspection systems already installed on the line. Nothing is replaced and no control loop is taken over.
Machines & PLC
Presses, welders, CNC, coaters, ovens, test rigs
Gauges & sensors
CMM, calipers, thermocouples, vibration, vision cameras
Line systems
MES work orders, AOI/SPI results, torque and leak testers
Edge gateway
Buffering, protocol translation, on-line inference
Responsibilities
- Normalise units, timestamps and subgroup boundaries at source
- Buffer locally for up to 14 days if the uplink drops
- Run latency-critical inference (vision, anomaly) without a round trip to cloud
Interfaces
- OPC UA
- MQTT / Sparkplug B
- Modbus TCP
- REST
- SQL / historian
- SFTP file drop
Runs on: Edge gateway on the plant network (virtual machine, industrial PC or container host)
Failure behaviour: If the uplink fails, inspection continues at the line and records sync in order once connectivity returns.
- Edge buffer
- 14 days
- Ingest rate per site
- 50k msg/s
- Vision inference
- Line rate
- Query on hot data
- < 1 s
Data flow
What happens between a measurement and a decision
Six stages with explicit latency budgets. Every stage keeps the context needed by the next one, which is why root cause does not require a manual data-joining exercise.
01Capture
10–50 ms · Machine, gauge, cameraInput
Tag change, measurement, image frame, test result
Output
Raw reading with asset and timestamp
Engineering note: Sampling plan and subgroup size follow your control plan, not a generic default.
Modules
Licence the capability the plant needs now
Modules share one data model, so adding a capability later never requires a second integration project or a parallel master data set.
Statistical process control
Drift visible while the process is still correctable
Live control charts and capability by characteristic, station, tool and shift, with configurable special-cause rule sets.
- Variable and attribute charts
- Cp, Cpk, Pp, Ppk
- Nelson / WE rules
- Subgroup and sampling plans
Inspection & vision
Consistent judgement on every part
Digital inspection plans for operators plus automated vision classification on existing camera and AOI hardware.
- Digital control plans
- Vision defect classes
- Reviewer feedback loop
- Gauge and CMM ingestion
NCR, CAPA & 8D
84% of actions closed on the first cycle
Nonconformance handling with containment, disposition, root cause, corrective action and verified effectiveness.
- Containment & disposition
- 5-why and Ishikawa
- 8D for customer escapes
- Effectiveness verification
Supplier quality
Incoming risk visible before the line stops
Incoming inspection, skip-lot rules, supplier scorecards and supplier corrective action in the same record chain.
- Incoming inspection plans
- PPM scorecards
- Skip-lot logic
- Supplier 8D portal
OEE & downtime
Quality losses separated from availability losses
Availability, performance and quality decomposition with reason-coded downtime tied to the same asset model.
- OEE decomposition
- Downtime Pareto
- Changeover analysis
- Shift comparison
Audit & compliance
Evidence retrieved in under a minute
Electronic records and signatures, retention policy, and export packs mapped to the standard being audited.
- Electronic signature
- Immutable audit trail
- Retention policies
- Standard-mapped export
Predictive quality
Escape risk scored before shipment
Batch and lot level risk models combining process data, genealogy, supplier history and inspection outcomes.
- Escape-risk scoring
- Drift forecasting
- Cause ranking
- What-if setpoint analysis
Dashboards & reporting
One scorecard from line to board
Role-scoped dashboards, scheduled management reports and cross-plant benchmarking on a single data model.
- Role-based views
- Scheduled reports
- Cross-plant benchmarking
- Export to BI tools
Deployment
From first connector to closed-loop control
Start with the data sources you already trust, then expand from pilot line to multi-plant quality operations without replacing your core systems.
- 1
Week 0–1
Connect
Integrate MES, ERP, historian, PLC, camera and gauge sources through native connectors, OPC UA, REST or secure file drop.
Data access confirmed
- 2
Week 1–2
Normalise
Map products, lines, lots, stations and critical characteristics into one governed quality model with genealogy intact.
Quality model signed off
- 3
Week 2–4
Monitor
Enable live SPC, capability tracking, AI defect analytics and alert routing for the agreed scope and owners.
Alerts in production
- 4
Week 4+
Close the loop
Route NCR, CAPA, supplier and audit actions with evidence and effectiveness checks, then measure against the baseline.
Results vs. baseline
Integration surface
Connectors run at the edge or in the cloud and never require write access to your systems of record unless you enable it explicitly.
- OPC UA
- MQTT
- SAP
- Siemens
- Rockwell
- Ignition
- Kepware
- REST API
AI engine
Six services, one rule: every output is explainable
The AI engine is decision support for qualified people. Each service records its inputs, model version, confidence and the human verdict, because an inspection decision you cannot explain is an audit finding waiting to happen.
Vision inspection
96% mean confidence on released classes
Classifies surface, weld, solder and assembly defects at line rate on your existing cameras, with every verdict stored alongside the image and model version.
Anomaly detection
Detects drift a median 42 min earlier
Multivariate models learn the normal operating envelope per asset and recipe, catching drift patterns that single-characteristic charts miss.
Root cause ranking
-62% investigation time
Correlates deviations against process parameters, material lots, tooling, operators and shifts, then ranks candidate causes by evidential strength.
Predictive quality
Escape risk scored per lot
Scores batches and supplier deliveries by escape probability so containment happens before dispatch rather than after a customer complaint.
Quality copilot
Grounded in your plant data only
Answers plain-language questions over inspection, process and CAPA history, and every answer cites the records it used.
Model governance
Full model lineage retained
Versioned models with drift monitoring, champion/challenger evaluation, approval gates and rollback — the audit trail an assessor expects.
Models are trained per tenant. Your process parameters, images and recipes are never used to train models served to another customer, and any released model can be withdrawn without losing the historical record of its decisions.
Model governance detailsProduct
Depth where quality decisions happen
Move from signal to owner to evidence without exporting data between inspection, analytics and action systems.
OEE
87.4%
Availability
93.1%
Performance
95.6%
Unplanned downtime
12min
OEE breakdown · shift B
target 85%OEE
+5.4 pt
Availability
+2.2 pt
Quality
FPY 98.2%
Downtime reasons · 7 days
- Tool change46 · 41%
- Material starve31 · 68%
- Sealing unit fault24 · 89%
- Quality hold12 · 100%
Machine health
4 assets- PRESS-0462%
Die temperature +6 °C vs. recipe
- WELD-0294%
Within nominal parameters
- AOI-1188%
Model v4.2 · 96% confidence
- SEAL-0741%
Bearing vibration trending up
Dashboards & analytics
The same numbers at every level of the organisation
Reporting is derived from one model rather than re-keyed per audience, so a plant review and a board pack cannot disagree about last month's scrap.
| Scope | Audience | Content | Cadence |
|---|---|---|---|
| Line | Operators, line quality | Current control state, open containment, inspection queue, first pass yield for the running order. | Live |
| Plant | Plant and quality managers | OEE, scrap cost, CAPA ageing, supplier holds and top defect families across every line in the site. | Live + shift summary |
| Region | Operations directors | Plant-to-plant benchmarking on identical characteristics, rollout status and improvement programme tracking. | Daily + monthly |
| Network | Executives, customer quality | Quality cost, escape rate, audit readiness and risk exposure consolidated across the manufacturing network. | Monthly + on demand |
Integrations
Connect what you already run
Connectors are read-first. Write-back to a system of record is enabled per integration, logged, and reversible — IT review usually starts and ends there.
Manufacturing execution
Work order, operation and genealogy context; optional write-back of hold and disposition.
- Siemens Opcenter
- Rockwell FactoryTalk
- AVEVA MES
- Critical Manufacturing
- Custom MES via REST
ERP & PLM
Material master, supplier, cost centre and specification revisions.
- SAP S/4HANA
- SAP ECC
- Oracle
- Microsoft Dynamics 365
- Siemens Teamcenter
Control & historian
Process parameters and asset state at sub-second resolution.
- OPC UA
- MQTT / Sparkplug B
- Modbus TCP
- AVEVA Historian
- Ignition
- Kepware
Inspection & metrology
Measurement results, images and pass/fail verdicts with gauge identity.
- Cognex
- Keyence
- Zeiss CMM
- Mitutoyo
- Koh Young AOI
- Test bench CSV/SQL
Cloud & analytics
Governed outbound feeds so corporate analytics reads the same numbers.
- Azure Data Lake
- Snowflake
- Databricks
- Power BI
- Tableau
Workflow & identity
Single sign-on, SCIM provisioning and task escalation into existing tooling.
- Microsoft Entra ID
- Okta
- Teams
- Slack
- ServiceNow
- Jira
Missing a system? The REST and streaming APIs cover anything a connector does, and our team will map the payload with you.
API overviewDeployment & security
Deployed to fit your data policy, not ours
Every option runs the same platform and the same models. The difference is where data lives and who operates the infrastructure.
Managed cloud
Regional tenant in EU, US or APAC with edge gateways per site. We operate the platform against a 99.95% SLA.
- Fastest time to value
- Automatic updates
- Regional data residency
Private cloud
Deployed into your own cloud subscription with your key management, network policy and observability stack.
- Customer-managed keys
- Your VPC and peering
- Change window control
On-premise / air-gapped
Full stack inside the plant network, including model serving, with offline update bundles and local retention.
- No egress required
- Offline model updates
- Local retention policy
Security programme
SOC 2 Type II aligned controls, role-based access scoped by site and product family, TLS 1.3 in transit, AES-256 at rest, customer-managed keys on enterprise agreements, immutable audit trail and an annual third-party penetration test. Validation support for 21 CFR Part 11 is available as part of an enterprise agreement.
Availability
- Monthly SLA99.95%
- RPO15 min
- RTO4 h
- Edge buffer14 days
Start the evaluation
Find out where quality drift is hiding in your plant.
Bring one line, one defect family or one audit workflow. We will map the available data sources, quantify the cost of the current detection delay, and show the fastest route to measurable control.
45-minute technical walkthrough
With a solution architect who knows manufacturing data, not a scripted demo.
NDA before any data review
We can assess feasibility from sample exports without production access.
Written pilot scope
Baseline metrics, success criteria and exit conditions agreed up front.