Founding 100 is open — 6 months of Pro free for the first 100 members  · 63 OF 100 CLAIMED (DEMO COUNT)
For Minitab users, Six Sigma Belts & quality engineers

General AI picks a statistical method that doesn’t fit the data about one time in three.

Rigor is the copilot that doesn't. You call the method first — then it checks every assumption in code, refuses the tests your data can't support, and computes every figure on a validated engine, never the language model. So the statistics behind your certification, your project, or your next decision actually hold up.

Free: 1 project · full coaching · no card · every number computed on a validated kernel — never by the AI
🏁 Founding 100: 6 months of Pro free + verification locked at $49 forever 37 SPOTS LEFT
64.8%
Best general-AI accuracy at selecting an applicable statistical method. Silently wrong ~1 in 3 times.
StatQA · NeurIPS 2024
−17%
Exam performance of students who leaned on an answer-vending AI tutor. Our predict-first loop inverts it.
Bastani et al., Wharton · 2024
100%
Of figures computed on a validated deterministic kernel and traceable to a logged run.
Architecture guarantee
What silent-wrong looks like

The p-value looked right. It nearly bought the wrong machine.

A Black Belt asked ChatGPT for a one-way ANOVA and got a clean, confident answer: p = 0.03 — significant, buy the equipment. The validated software said p = 0.21 — not significant. The AI hadn't calculated anything. It predicted what a plausible answer looks like.

General AI · one-way ANOVA
p = 0.03
"Significant — buy the equipment." Wrong.
Rigor's validated kernel
p = 0.21
Not significant. The purchase doesn't hold up.
source: AIGPE · Six Sigma reality check · 2026
32.5%
How often ChatGPT produced the correct statistical values in ordinary, plain-language use.
Not method selection — the numbers themselves. On independent t-tests and one-way ANOVA under normal prompting it scored near zero. The prose stayed perfect throughout.
source: JMIR · 2025 · PMC11845875

This is the failure Rigor removes by design: it commits you to a method first, checks the assumptions in code, and computes every figure on the kernel — so the number you hand an examiner is one you can defend, not one that merely sounds right.

rigor://vera · try it before you sign up

Ask for an analysis.
Watch Vera run it — or refuse it.

Type a real request — or tap one below. Vera commits to the method she'd defend, runs it on a sample dataset — chart and computed result — and refuses the ones the data can't support. Every figure comes off the kernel, never the model.

Vera · methodology router · kernel 1.3.2
demo mode
V
Hi — I'm Vera. Tell me what you're trying to find out about your process, and I'll pick the method I'd stake my name on. Or start with one of these:
Scripted preview of the real router · the live engine checks assumptions on your actual data
rigor://architecture · click a stage

The judgment layer, stage by stage.

Six gates between your raw data and a claim you can defend. The one in red is the one general AI skips.

01
Your data
02
You predict
03
The wall
04
Kernel computes
05
Storyboard
06
Verified
How it works

Calibrate. Predict. Prove.

No tool menus, no 40-hour learning curve. One project, one methodology brain, and an interface that makes you better instead of dependent.

01 · 4 minutes

Calibrate

A short scenario assessment — five real situations — measures how you think, not what you've memorized. Rigor configures itself — coaching depth, mode, emphasis — and your answers become your measured baseline.

"Is this process stable?" → you choose
skill read: charts strong · hypothesis tests → coaching here
02 · The core loop

Predict-first bench

Before any analysis runs, you commit to your method call — then Rigor reveals its selection, the checked assumptions, and the why. Correct calls build your evidence log; misses become coaching.

your call: A → reveal: 2-sample t · checks 3/3 PASS
run a-41f · kernel 1.3.2 · logged unaided
03 · The deliverable

Verified storyboard

Your DMAIC storyboard assembles itself as you work — charts to SPC convention, every figure provenance-stamped. Export, then have the independent examiner verify it.

8 slides · provenance appendix
VERIFIED · VER-2026-08C41 · registry QR
Your learning, measured

An AI coach that knows
why you miss.

Calibration starts with a short skills assessment — and the gaps it finds become your coaching emphasis, automatically. From there, every method call you make updates a live mastery estimate per skill, tracked down the side of your workbench as you complete real work.

  • Mastery, not gold stars — each skill carries a probability estimate that moves with every prediction you make
  • Misconception diagnosis — the AI clusters your misses into named patterns ("independent vs paired," "mean-comparison trap") and coaches the pattern, not the incident
  • Decay modeling — skills fade without practice; Rigor notices, and queues 60-second spaced drills before the exam does
  • Exam readiness — everything rolls up against the actual certification blueprint, so "am I ready?" has a number

Answer-vending AI tutors made students score 17% worse on the unassisted exam. Rigor is built to move that number the other way — and to show you it moving.

Skills · AI-tracked ● LIVE ON YOUR WORKBENCH
Charts & stability0.86
Capability thinking0.74
Hypothesis testing0.57
◉ Coaching focus · predict-first slows down here
MSA / Gage R&R0.61
Idle 12 days · fading — 60-sec refresh queued
AI coach read Your misses cluster on one pattern — M-03 · independent vs paired — resolved after 3 exposures. Next gap: practical vs statistical significance. I'll probe it on your capability study.
Exam readiness · ICGB blueprint74%
Analyze domain closing — strongest gain this week
Why trust it

Four walls between the AI and your project.

Minitab's AI summarizes results after you've already chosen the analysis. ChatGPT will compute capability on an unstable process without blinking. Rigor is built the other way around.

Wall 01 · Computation

The AI never computes

Every statistic runs on a deterministic kernel validated against NIST and Minitab reference outputs. LLM output physically cannot contain a computed number — a post-processor enforces it.

Wall 02 · Judgment

The AI never freelances methods

Method selection runs through methodology playbooks and a DMAIC state machine. Preconditions are checked in code — an analysis whose assumptions fail doesn't run.

Wall 03 · Evidence

Everything is traceable

Every figure carries its run ID, dataset hash, kernel version, and your decision trail. Hand it to a grader, a boss, or an auditor — "how did you get this?" is answered on the page.

Wall 04 · Causal chain

The AI never breaks the logic

Your project is one causal argument — Y=f(x), from problem to proof. Rigor keeps the thread: solutions must trace to confirmed causes, "improved" is a refused claim until post-change data proves it, and controls attach to the causes you actually verified.

REFUSED —

"Run capability on this data."  Your process shows a special cause on July 11 and fails the stability check. Capability on an unstable process is meaningless — let's chart it, find the cause, and then measure capability. That refusal is the product.

The credential

Don't just certify.
Get Rigor Verified.

Employers know certificates can be bought. So we built the thing they can check: an independent adversarial examiner re-computes every claim in your storyboard from your raw data — then issues a tamper-evident record any recruiter can verify in one click.

  • Every figure recomputed on the kernel and diffed against your deck
  • Method selection re-derived; assumptions re-checked; soft conclusions flagged
  • Pass → Verified stamp, QR, and a public registry record
  • Fail → a defect list and a free re-review, not a dead end
  • Built your project in Minitab or Excel? Verify it anyway — any tool
  • Sustainment re-check at 90 days — the examiner re-runs your frozen Control contract on fresh data (live connector or attested upload) and adds a second stamp: gains sustained
  • No stamp is ever a failure mark — absence just reads "not re-verified"
● RIGOR VERIFIED
VER-2026-08C41
ProjectGB-001 · Fill-weight variation
Examiner / kernelv1.0 · 1.3.2
Figures recomputed14 / 14 match
Candidate decisions21 logged · 16 correct unaided
Sustainment● SUSTAINED · 90 DAYS · CONNECTED DATA
RecordTamper-evident · anchored
Recruiters: verify any record free at the registry
Founding members

Early belief, permanently rewarded.

The ladder steps down as we grow — earlier members get more, forever. We give away time and locked prices, never the credential: verification always costs something, because a registry only matters if its stamp can't be handed out. And the counters are real — a company named Rigor doesn't do fake scarcity.

● Open now

Founding 100

The first hundred believers
  • 6 months of Pro, free (worth $174)
  • Project Pass included
  • Verification locked at $49 forever
  • Founding badge on your registry record
  • Direct line to the founder — you shape v1
CLAIMED63 / 100
Opens at 100

Charter 500

Members 101–500
  • 3 months of Pro, free (worth $87)
  • 50% off the Project Pass
  • Verification locked at $79 forever
  • Charter badge on your registry record
  • Founder line — Founding 100 only
CLAIMED0 / 400
Opens at 500

Early 1000

Members 501–1,000
  • 1 month of Pro, free
  • Early-adopter pricing locked at launch rates
  • Early badge on your registry record
  • Then: standard free tier for everyone, forever
CLAIMED
Not sure yet? Take the free 4-minute skills check first — see how your statistical judgment ranks against 511 quality professionals. Start →

Reserve your founding spot

Rigor isn't open to the public yet — we're onboarding the Founding 100 first. Leave your details and your spot is held, price locked, no card. Loading the live count…

No card · one claim per person · unsubscribe anytime · we never share your email
Counters are live and honest — a company named Rigor doesn't do fake scarcity · one claim per person · offers stack with the AIGPE course code
Pricing

Do the work free. Own it for $49.
Prove it for $99.

Your exam costs $295–438. Your prep course cost hundreds more. Rigor is the cheapest part of your certification — and the only part that produces checkable evidence.

🔒 Not open yet. These are launch prices — join the Founding 100 to lock them forever (plus 6 months of Pro free) before we open to the public.
Start here

Free

$0
forever
  • Full copilot & predict-first coaching
  • 1 active project · ~20 analyses
  • Unlimited practice datasets
  • Watermarked export
Reserve at launch
Most popular

Project Pass

$49
one-time · 90 days
  • Everything in Free, uncapped
  • Clean .pptx / PDF export
  • Provenance appendix + Locked mode
  • Verification at $79
Reserve at launch
Professionals

Pro

$29
per month · cancel anytime
  • Unlimited projects
  • Auto mode (earned via calibration)
  • Priority kernel runs
  • Verification at $79
Reserve at launch
The credential

Verification

$99
per project · $79 with Pass/Pro · BYO $129
  • Independent adversarial examiner
  • Verified stamp + registry record
  • Free re-review on fail
  • Any tool — Minitab & Excel welcome
  • 90-day sustainment stamp: +$39 prepaid · $59 later
Reserve at launch
ASQ CSSGB exam$438 / $338
IASSC ICGB exam$295
Exam retake$269
GB prep course$500–2,500
Minitab seat$2,394 / yr
Your data stays yours. Encrypted, never used to train models, deletable anytime. Regional pricing in 190+ countries at checkout. Students & course partners: your instructor's code preloads practice datasets.

Quality taught the world to verify before trusting.

Time our tools did the same. Five pilot spots open this month.

Join the founding waitlist