Blue Star Labs · The Plant Score · Demo

One walkthrough. One score.

Can your plant See its machines, Trace its answers, Adopt its own best fixes, and Retain what its veterans know? Four pillars. One number. Red or green.

This is a demo scored on Ridgeline Instruments — a fictional plant. Every number below is synthetic, and every one of them recomputes by hand. No client data exists anywhere on this page.
Skim the one-pager — print-ready Or scroll the full case study ↓

The whole assessment fits on one page — the way a plant reads anything. Everything below is the drill-down.

The scoreboard

Ridgeline Instruments · Plant Score

Electronics manufacturer — two SMT lines, four assembly cells, three shifts, fourteen critical machines. The flagship is the Horizon industrial printer line. Scored on a half-day facilitated walkthrough plus one completed 30-day sprint.

Every value on this page carries a tag: how we know it. measured assessed self-reported not yet measured

The four pillars

See it. Trace it. Adopt it. Retain it.

Twelve components, each scored by a written rubric an engineer can audit — graded questions, timed exercises, and evidence pulled on the floor. Nothing here is a survey answer wearing a score costume: if we didn't see it, it says so.

The rule that makes it honest: anything unmeasured leaves the math — numerator and denominator. It is never scored zero, never defaulted, never guessed. A smaller map, not a penalty.

The escalation board

Red or green. On track or not.

Every critical finding goes on the board and gets a lane, an owner, and a day counter — the way a plant actually manages a problem. The four steps are the same STAR loop: See it (make it visible) → Trace it (root cause) → Adopt it (standardize the fix) → Retain it (prove it holds). Management reads it in ten seconds.

An escalation leaves the board one way: RETAIN verified — the fix held, in control, with the receipts. Not when the meeting ends. When the process proves it. ESC-03 above just earned its exit at this week's review.

The mirror

“About 80.”

At the walkthrough, three people wrote down what they believed Line 4 runs at. The cards were locked before anyone measured anything. Thirty days of real machine data later:

“Your effectiveness is 54 — and your Plant Score went up when you measured it.”

The score rewards seeing, not pretty results. It can only go up by installing sight, trace, adoption, and retention — never by hiding a bad number. That's the whole point.

How a score hardens

Testimony → assessment → measurement

The same plant, three stages. A phone call earns a band. A walkthrough earns a number. A sprint replaces belief with 30 days of machine data — one pillar at a time.

Grading testimony cost a band — measuring won it back. The score moves only when knowledge does.

The math

Your engineer is welcome to audit this.

One formula, printed weights, no black box. Every component score traces to a written rubric, a timed exercise, or 30 days of machine data — and the whole thing recomputes by hand from the signed worksheet.

PlantScore   = round( Σ(component × weight) ÷ Σ weights )   over SCORED components only
pillar_score = same formula, over that pillar's SCORED components

This plant:  4,155 ÷ 95 = 43.7 → 44   (one component unmeasured → its weight 5 leaves the math)

Weights — equal by design in v1

Each pillar carries 25 of 100. We have no field evidence yet that one pillar predicts more than another — so we refuse to pretend we do. Weights are versioned; when the evidence exists, the version changes and every re-score says so.

weights: estar-2026-07-w1 · bands: estar-2026-07-b1 · definition: estar_v1

The bands

Iron rules

  • People write the rubrics; the rubrics own the numbers. AI never invents or inflates one.
  • Unmeasured leaves the math. Never zero, never a default.
  • Every value says how we know it — and self-reported never outranks measured.
  • It scores the operating system, never the results — so honesty is always the winning move.

The pipeline cost sheet

What does the Friday report cost?

At Ridgeline, the weekly report is hand-built: three engineers re-keying audit workbooks every day, a Friday assembly, a Monday argument about whose number is right. We timed it at the same walkthrough: 18.5 engineer-hours a week — $55,500 a year at a printed $60/hr loaded rate. Five cost lines — build it, check it, fix it, defend it, change it — each tagged with how we know it, and each recomputing by hand from the timing worksheet.

The line we refuse to price: decisions made late — or made twice — on numbers nobody fully trusts. The biggest cost in the building stays out of the total, because we didn't measure it. Even when pricing it would make the pitch better.
See the pipeline cost sheet — print-ready

Page 1 of the report is the score; page 2 is this sheet. The score says whether your system can see itself — the sheet says what the current one costs in payroll.

Want your real number?

Half a day on your floor. Twenty-two questions with written grading anchors, four exercises your own people run, and a signed read-back before we leave. One fixed price, put in writing before you say yes. Scored by hand, on paper you keep. Your data never leaves your building — we walk out with derived scores and finding titles, never your files, your photos, or your data.

What you actually get: a report whose page 1 is this exact one-pager — made to be printed and pinned to the board — with the drill-down behind it. Nothing lands on your team's plate: your people hold the stopwatch for an afternoon, and we leave you a scored map plus fixes you can buy or do yourself.

Ask about a baseline walkthrough

Or just email john@bluestarcentral.com. And if you want to see where the clean data behind the SEE pillar comes from, read Case File 01: The Daily Audit Problem.