Halia v1.0 · the detection engine inside Instirio NEW Find every leak your dashboards miss.
37 detectors run on your live order stream. Each finds a different pattern your dashboards miss, ranked by dollar impact.
Every finding ships with a dollar figure, a confidence score, and the math behind it.
37 detectors. Halia orchestrates them.
Every order event flows into the engine, which fans across three tiers, structural, behavioral, operational, then ranks findings by dollar impact and confidence.
How does Halia combine ML, statistics, and graph reasoning?
It isn’t ‘an AI’. It’s a deliberate stack, statistical baselines, machine learning, graph reasoning, continuous monitoring, each handling the part of the problem it’s best at.
Calibrated baselines
Every metric, transition, and actor gets a continuously-updated baseline calibrated to your store. Halia flags meaningful deviations from your normal, not someone else’s.
Pattern intelligence
Trained models distinguish system, carrier, and human activity, surface emerging process variants, and detect time-series anomalies that simple thresholds miss.
Process conformance
Halia models your order flow as a graph and compares it to the canonical lifecycle. Bottlenecks, rework loops, and missing steps surface naturally.
Continuous monitoring
Every detector runs continuously against each metric’s expected range. When a problem you fixed starts returning, you’ll know within an hour, not at the next quarterly review.
Severity & impact scoring
Findings are ranked by a multi-factor severity score that blends delay, dollar impact, and volume. So the issues at the top are the ones worth fixing first.
Provenance on every number
Every finding ships with a confidence score derived from sample size and signal strength. Numbers without provenance are guesses with attitude. And we don’t ship those.
What industry standards does Halia align with?
No invented metrics. We use the language Fortune 500 supply-chain leaders, IEEE researchers, and ISO auditors already speak, calibrated for SMB scale.
4-factor POR multiplication, 3-phase cycle decomposition (Source / Make / Deliver), 5-category Cost-to-Serve breakdown.
Every event in Instirio is XES-aligned with. Export to OCEL or XES for audit, research, or import into other process-mining tools.
16 automated health checks (8 P0 every sync, 8 P1 hourly) validate the four ISO/IEC 25012 characteristics continuously.
Halia is always running.
Reads every order event the moment it happens. No batch jobs, no overnight reports, no Monday-morning surprises.
- Real-time ingestion, events from Shopify, WooCommerce, Stripe arrive in milliseconds.
- Continuous baselines. Every metric is monitored continuously against your store’s normal range. No “wait a quarter to see drift.”
- Tenant-fair detection. The engine balances load across all customers so your detection cycle stays predictable.
- Pulse notifications, drift events appear on your dashboard within seconds.
Halia event log · sample teamSample event log. Real timestamps update continuously inside Instirio.
How does Halia calibrate confidence on every finding?
Every finding ships with a confidence score from sample size, signal strength, and detector overlap. Numbers without provenance are guesses with attitude.
Bottleneck
2,090 affected orders
Drift
30-day baseline · 412 orders
Time pattern
14-day window · 168 hourly buckets
Stuck orders
463 affected · no batch cap
What makes Halia different from a generic LLM?
| Capability | Halia (Instirio) | Generic LLM “AI” | Spreadsheet rules | Celonis copilot |
|---|---|---|---|---|
| Standards-grounded math (SCOR / XES / ISO) | ✓ | – | – | ✓ |
| Auditable provenance per finding | ✓ | – | ✓ | ✓ |
| Multi-method (stats + ML + graph) | ✓ | ~ | – | ✓ |
| Confidence calibration | ✓ | – | – | ✓ |
| Drift / regression detection | ✓ | – | – | ~ |
| Hallucination risk | None | High | None | Low |
| Self-validating health checks | 16 automated | – | – | Partial |
| Cost | $0–$149/mo | ~$20/mo | Free | $100K+/yr |
Where Halia’s math is anchored.
It is not a black box. Every detector’s methodology, every metric formula, and every threshold is grounded in public, peer-reviewed standards.
- ASCM SCOR Digital Standard. The Association for Supply Chain Management’s framework.Halia implements RL.1.1 (Perfect Order Rate), RS.1.1 (Cycle Time), and CO.1.1 (Cost to Serve). ASCM.org →
- IEEE XES 1849. The IEEE process-mining event log standard. EveryHalia event uses XES-compliant fields (case_id, activity, timestamp, resource, lifecycle). IEEE 1849 standard →
- ISO/IEC 25012:2008. The international data quality model.Halia’s 16 health checks (8 P0, 8 P1) validate the four ISO characteristics: Completeness, Accuracy, Consistency, Traceability. ISO.org →
- van der Aalst on process mining. The academic foundation of process mining.Halia’s conformance checking and variant analysis use techniques from his open-access textbook. Springer →
Run Halia on your data.
Free for 500 orders/mo. 5-minute setup. Standards-aligned with from the first finding.
Start free →