Three projects, worked end to end.On the product’s own demo data.

In the demo, Acme’s improvement team is Elena Torres (admin), Sandra Kirchner (the OE manager who reviews tollgates), David Rodriguez (an executive with read-only access) and three practitioners: Priya Nadar, Marco Silva and Nathan Chen. The organization has 25 seats, seven of them filled counting its owner, and a five-day service level for tollgate reviews.

Its problems are the everyday ones of a plant. Packaging Line 3 defects averaged 4.2% over Q1, up from 2.8% in Q4, driving about $12,000 a month in scrap. Changeovers averaged 47 minutes against a 22-minute benchmark. The wrong SKU was picked on 2.1% of warehouse orders. Boilers ran at 71% efficiency against an 82% benchmark. The forklift fleet was down 18% of the time. Sales orders were reworked 14% of the time because of missing or invalid data.

The demo shows the program at one moment, with each project at a different point. Forklift Downtime Reduction is the one that has been all the way through, so it is the one to follow phase by phase. Elena Torres led it with Marco Silva and Nathan Chen, and Sandra Kirchner reviewed every gate.

1. Step 01

   Define

   A project charter from DocuForge: downtime averaging 18% against an 8% target, with a goal to halve it and save about $85,000 a year in rentals and delays.

2. Step 02

   Measure

   A baseline study in DataForge confirmed 18% average downtime.

3. Step 03

   Analyze

   A DataForge Pareto showed four causes account for 81% of the downtime.

4. Step 04

   Improve

   A designed experiment (DOE) in DataForge to optimize the preventive-maintenance interval.

5. Step 05

   Control

   An I-MR control chart holding downtime at 6.4% with no out-of-control points, and an ImpactMatrix record of the realized impact.

• Project charters from DocuForge on every project, plus a SIPOC of the packaging process and a data collection plan for Line 3.
• A Line 3 baseline capability study (Cpk 0.72, not yet capable) and a Gage R&R on the torque gauge at 4.5% of study variation, acceptable, so the baseline describes the process and not the instrument.
• Pareto charts that narrowed changeover time to cleaning, tool changes and calibration (72% of the time) and forklift downtime to four causes (81%).
• A regression of boiler efficiency on load (R² 0.78) and a designed experiment on burner tuning.
• A fishbone, a five-whys chain and a mistake-proofing design from the Lean tools; the order-entry fix is form validation.
• Control charts: an I-MR chart holding forklift downtime at 6.4%, and a p-chart with order-entry rework trending to 4%.
• An ImpactMatrix return for the forklift project: $92,000 a year against an $18,000 implementation, paying back in 2.3 months.

Forklift Downtime Reduction closed with all 5 gates approved by Sandra Kirchner, none of them by Elena, who submitted them. The team projected $85,000 a year and claimed $92,000. The organization’s owner, who did not lead the project, validated $88,000 with a note: verified against rental invoices and maintenance logs, and $4,000 of the claim was one-time, not annual.

Across the five open projects, $560,000 of savings is projected, and the results view keeps it apart from the $92,000 claimed and the $88,000 validated, so nobody adds an estimate to a result.

The delivery board shows what needs attention without anyone compiling it: the Line 3 Measure gate is waiting for Sandra’s review; Warehouse Picking Errors has had no activity for 14 days and is flagged at risk; Order Entry Rework is past its target date; and Boiler Energy Optimization’s Improve phase was due ten days ago, so its plan shows as slipped. The idea queue holds four ideas, one at each stage: new, scored, approved and declined.

• Validate against source documents.

  The $4,000 difference was a one-time saving counted as annual. It was found because someone other than the lead checked the invoices.

• Measure the gauge before the process.

  With the Gage R&R at 4.5%, the Cpk of 0.72 is a fact about Line 3, and the team can act on it.

• Let the board run the status meeting.

  A waiting gate, a quiet project, a late one and a slipped plan are four decisions, and the board lists them.

Prof. Anita Deshmukh teaches QE 401, Quality Engineering, at State University in Fall 2026, on the Green Belt track. She wants every student to finish with a project they could defend, not only a quiz score, and she is one person reading every submission.

Two partner organizations have published project briefs on the course. Riverside Beverages: bottling Line 3 changeovers average 47 minutes against a 25-minute target, costing about nine hours of capacity a week since new SKUs arrived in March. The State University Library: one in five help-desk tickets is reopened within a week, and wait times double in the first week of term.

The term follows one student, Sofia Marchetti, the student the university video follows too, from the course form to her Measure gate.

1. Step 01

   Set up

   One form: course name, Green Belt track, term, start and end dates, and a due date for each of the 5 capstone gates. The course gets the join code QE401-F26.

2. Step 02

   Join

   Students join with the code. Each is enrolled in the Green Belt lessons and given their own DMAIC capstone in the university’s workspace.

3. Step 03

   Take a brief

   Sofia takes the bottling-line brief. Its problem becomes her problem statement, and her goal is to cut Line 3 changeovers from 47 to 25 minutes by the end of term, held for three weeks.

4. Step 04

   Measure

   Her Measure evidence is a DataForge histogram of 34 changeovers: mean 47.1 minutes, standard deviation 6.8, from 33 to 62. She submits the Measure gate.

5. Step 05

   Grade

   Submitted gates from the whole course arrive in one queue with an AI pre-read against the rubric. The professor approves or returns each one, with feedback.

• A roster with each student’s lesson progress, capstone gates and standing, beside the join code and the course dates.
• Briefs ranked by a priority score from impact, effort, confidence and fit, the same scoring the enterprise idea queue uses.
• A gradebook of every student’s 5 gates, an AI pre-read and drafted feedback for a submission under review, and an export to CSV.
• An advising view of every capstone’s gates and standing, the cohort by phase, and a suggested intervention.
• Course certificates that name the course, the term and the supervising professor.

Mid-term in the demo, three gates are waiting in the professor’s queue, each with its evidence attached: Liam O’Connor’s Define, Sofia Marchetti’s Measure and Jordan Blake’s Analyze. Two students, Lucas Meyer and Nina Petrov, are 15% and 5% through the lessons and have gone quiet, and the course flags them as at risk.

Maya Krishnan has finished. All 5 of her gates were approved by Prof. Deshmukh, her capstone passed, and State University issued her course certificate. She also took the Certification Upgrade and passed its exam, so she holds the separate SigmaForge Professional Certification, GB-2026-0001, which anyone can check on the verification page.

The university film runs the same course on to the end of term, where Sofia finishes, passes the exam and holds GB-2026-0003.

• Real briefs make real projects.

  A capstone that starts from a partner’s problem arrives with a baseline and a target already stated.

• The pre-read saves reading, not judgment.

  The AI reads each submission against the rubric first. The grade is still the professor’s.

• Keep the course and the certification apart.

  The course certificate records what the university assessed. The professional certification needs its own exam.