# Terminus M4 — Quality decisioning

> A temperature excursion assessed against the product's own stability model, in under a minute, with the trail an inspector will ask for.

- Where it stands: Commercial
- Quality · Synlogica
- Canonical: https://synlogica.ai/applications/terminus-m4/
- Language: en-US

Terminus M4 reads the logger export and the stability documents, models what the excursion actually cost in shelf life, and returns a RELEASE / CONDITIONAL / HOLD recommendation with a stated confidence interval. The authorized reviewer makes the binding call; the sealed decision package records how.

## The problem

A two-day excursion review is not slow because the maths is hard. It is slow because the evidence is scattered across a logger CSV, a stability report and a reviewer's spreadsheet, and because every reviewer reconstructs the assessment differently. The result is defensible only as long as the person who did it is still in the building.

## What it does

1. Screens every input before a number is computed — gaps, outliers, unknown logger layouts — and pins each file into the audit trail.
2. Turns the temperature profile into MKT and Arrhenius k(T) against the product's own stability model, then runs Monte-Carlo to state a 90% confidence interval rather than a point estimate.
3. Returns an advisory recommendation with the exact rule and model versions that produced it, sealed as a decision package the QA reviewer signs.

## Key capabilities

- **Sensor import for any logger layout** — Logger exports are parsed and screened before use — gaps, outliers, out-of-range readings, unknown formats confirmed by a person (ALCOA+). Every file is pinned into the audit trail.
- **Frozen kinetics, pinned by hash** — MKT and Arrhenius k(T) run against the product's own stability model. The mathematics is GxP-frozen and pinned by a SHA-256 hash, so a result can be reproduced bit-for-bit years later.
- **Monte-Carlo confidence, not a point estimate** — Thousands of runs turn the kinetics into a probability with a stated 90% confidence interval and the data-quality flags behind it.
- **RELEASE / CONDITIONAL / HOLD with QP signature** — The advisory verdict is mapped by QP-approved rules; the binding disposition is signed per series by the Qualified Person with password-bound e-signature.
- **Excursion Case Management** — Customer excursion cases with roles for customer service, QA, QP and the stability expert, due dates, closure rationale and an append-only audit log.

## Who it is for

QA and QP teams that assess cold-chain and ambient excursions every week, and the validation teams that have to defend the method at inspection. Advisory by design: batch disposition stays with the authorized human, recorded in your QMS.

## Where it stands

Commercial — available under licence today. Onboarding starts with a working session on your own data, where the configuration baseline and the shared-responsibility split are agreed before anything goes live.

## The same foundation as every Synlogica application

Every application in the portfolio carries the same controls, so an inspector who has seen one knows how to read the next.

- Versioned records — a record can be superseded, never silently altered.
- Electronic signature on every controlled step.
- An audit trail an inspector can follow, hash-chained end to end.
- Runs on your own tenant, hosted in the EU.

## Read more

- [From a 2-day excursion review to a sub-60-second decision package](https://synlogica.ai/resources/excursion-decision-package/)
- [Arrhenius shelf-life modelling — practical guide for QA teams](https://synlogica.ai/resources/arrhenius-shelf-life/)
- [How to calculate MKT (Mean Kinetic Temperature) — with a worked example](https://synlogica.ai/resources/how-to-calculate-mkt/)

## Talk to us

Bring a real case — a recent excursion, an invoice you doubted, a procedure that just changed — and we walk it through together. If it is not a problem for software, we will say so.

- Book a 30-minute call: https://calendly.com/adam-karpinski-synlogica/30min
- Email: adam.karpinski@synlogica.ai
- Privacy Policy: https://synlogica.ai/privacy/