Assay

Specification-drift intelligence

Sign in
Stainless steel process vessels and pipework in a dairy plant at dusk

Works with your plant, your laboratory and your ERP

Catch the batch while it can still be saved.

Your plant already holds the answer: control-system trends, your laboratory's results and the ERP batch record, sitting in three systems that were never designed to be joined. Assay joins them, learns what a good run looks like on your own line, and gives the operator the likely outcome with enough lead time to act. The laboratory result remains the measured truth; Assay simply stops the plant waiting on it in silence.

4.2 h
Median lead time gained before the result is confirmed
£184k
Recovered by Assay plants this year
97%
Of laboratory samples resolved to their batch

Line 2 · Batch NV-24188

Fat content

Marginal
3.42% w/wband 3.36 – 3.48
3.20 lower spec3.60 upper spec

Why it is moving

Jacket temperature has averaged 1.8 °C above the golden-batch mean over the first half of the run.

Separator feed pressure settled 9 minutes later than usual.

Suggested action: hold jacket setpoint at 68 °C for the remainder of the run.

What makes it trustworthy

01

It joins what nobody else will

Exact identifiers first, then normalised codes, then near matches, then sample time minus your laboratory lag. Where two answers are possible, Assay refuses to guess and asks.

02

It predicts a range, never a number

Every prediction carries an uncertainty band and a conformance state in words. A model that cannot beat your own historical average is never shown on the line.

03

It states the reason in your words

Not a coefficient. “Jacket temperature has averaged 1.8 °C above the golden-batch mean over the first half of the run.”

Start before you connect anything

A retrospective analysis of your own last twelve months.

Give Assay historical process exports, your laboratory's results and the ERP batch records. It joins them, quantifies what yield loss, out-of-specification product and giveaway actually cost, and shows how many of those failures were already visible in the process signal. No installation, no integration project, no risk to production.

See the retrospective analysis

Yield loss

Shortfall against your own best-decile run, costed at material plus conversion.

Out of specification

Failing batches costed at material, conversion and disposal.

Giveaway

Material handed over above target on batches that passed perfectly well.

What Assay is not

Assay is production process intelligence. It sits alongside the people and systems a plant already trusts, and it is deliberately narrow about what it will do.

Not a replacement for your laboratory

It is not a LIMS and it does not test anything. It reads the laboratory's results where you have a system, or the quality manager's spreadsheet where you do not, and treats every result as the measured truth — with its method uncertainty, its sampling regime and its reporting lag modelled properly.

Not in control of your plant

There is no write-back to control systems. Assay advises, the operator decides, and no automated action is ever taken on the line. That is a deliberate safety position and it is why plants are willing to let it in.

Not a job for an automation team

OPC UA where you have it, and the file-based historian and laboratory exports mid-market quality departments actually use where you do not. Configuration is done in the interface by the quality manager, not by a systems integrator.

Three screens, three jobs

The operator, the quality manager and the director each need a different truth from the same data. Assay does not make any of them read someone else's dashboard.

Operator

Line

Readable at two metres. One batch, the predicted result, the band, and what to do about it.

Quality manager

Batches & models

Every batch, how it was joined, and whether each model has earned its place on the line.

Director

Value

Scrap avoided, giveaway reduced, yield recovered — each entry traceable to a batch and an action.

Already working on a real plant.

Northvale Creamery uses Assay with two years of batches, trained models and a populated value ledger. Nothing is mocked.

Open Assay
The Assay app icon

Install Assay on the plant tablet.

Assay installs as an app on a tablet, phone or desktop. It opens full-screen straight at sign-in — no landing page, no browser furniture — and every screen behaves exactly as it does here: the live line board, batch records, retrospective analysis, model health and the value ledger.

On iPad or iPhone: Share → Add to Home Screen.