Proof & benchmarks
Measured where possible. Modeled where necessary. Labeled either way.
See what TwinEdge has measured, what we have modeled, and how we validate results at a customer site. Every claim is presented with its evidence type and limitations.
Measured
Results observed in a stated local test, with the test limits kept visible.
Modeled
Illustrative outputs from representative physics models—not customer results or guarantees.
Site-validated
Outcomes measured against an agreed customer baseline during a scoped evaluation.
Measured evidence
TwinEdge OS local edge benchmark
In a 60-second local lab probe at the 20,000-tag tier and a 100 ms publish interval, the TwinEdge OS C/open62541 data plane stored the stream in Apache IoTDB with no observed sequence gaps or replay backlog.
20,333.3
stored readings/s
100 ms
p95 source-to-ingest
1.000
store ratio
0
sequence gaps
How to interpret this result
- This is a local engineering benchmark, not a production-sizing commitment or hardware certification.
- The Siemens IIH public 20K figure is useful context, but its live-value update metric is not identical to TwinEdge source-to-ingest latency and confirmed storage acceptance.
- Production sizing requires the same test on the target ARM64 gateway or x86 industrial PC, including a longer soak and power-loss recovery.
Modeled projections
Representative outcomes—not observed customer results
These figures are illustrative outputs from TwinEdge physics-based models on representative assets. They show the kinds of outcomes the platform is designed to evaluate; they are not universal expectations or guarantees.
22%
less energy spend
Modeled for representative pumps, fans, and chillers operating closer to their efficient range.
3.4×
longer between failures
Modeled from earlier detection of progressive equipment drift and cause-based intervention.
73%
fewer inspection trips
Modeled where continuous online condition assessment can replace eligible routine rounds.
A site-specific estimate requires actual operating data, asset condition, process constraints, and current operating policy. Request a scoped evaluation to establish a defensible baseline.
Site validation
How a TwinEdge evaluation is measured
We begin with a narrow operational question and agree on success criteria before measuring results. Customer outcomes remain customer-specific.
Step 1
Agree on the baseline
Define the comparison period, assets or sites in scope, and the operational outcome to measure.
Step 2
Connect a bounded workflow
Connect your source systems and operating workflows from recommendation through closeout.
Step 3
Confirm controls
Document deployment constraints, approval roles, safety boundaries, and any permitted writeback.
Step 4
Measure before expanding
Compare results with the agreed baseline, record limitations, and expand only when the evidence supports it.
Governance boundaries
Recommendations remain governed by people and policy.
TwinEdge agents are recommendation-first and read-only by default, with dry-run, approval, audit, and replay controls. Safety PLC and SIS remain authoritative. Mobile diagnostics provide screening evidence, not certified scientific measurement or safety clearance.
Evaluate TwinEdge
Establish proof against your own operating baseline.
Explore the complete TwinEdge AI package for your operation. Launch subscriptions start at $3,750/month billed annually; final scope depends on deployment, integrations, compute, and support.