The Partner Certification Rubric
Version: 1.0 · 2026-09-19 What this answers: every line a candidate for certification is marked against, on both paths, with each line's standing as mandatory, flexible or locked, and what a pass, a resubmission and a fail take. Who this is for: candidates, certified partners, the house reviewer who marks, and any client who wants to read what their partner was held to. Status: published in full, which is a settled decision rather than a courtesy. D10 of The Certified Partner Program: "Publish it in full. A rubric a candidate cannot read is a rubric they cannot be held to, and the locked lines are exactly the ones we want a partner quoting to their own client."
Why this document exists separately. D11 settles that a badge is dated and tied to the version of the rubric it was earned against. A rubric carried inside a larger design document takes that document's version number, so a badge would have had to cite a version that moves every time an unrelated economics table is re-run. This document carries its own version and changes only when a line changes, so the version a badge cites means what it says. The Certified Partner Program points here and no longer carries a second copy of these tables.
1. How a line is marked
Marked the way a recipe checklist is marked. Each line is one of three things, and the three are not degrees of the same thing.
| Standing | What it means | What a failure of it costs |
|---|---|---|
| Locked | Nobody may relax this line for any candidate, in any circumstance, for any result elsewhere | A hard fail on its own. Never traded against a strong result anywhere else, and a resubmission is not the right answer to it |
| Mandatory | The line has to pass for the submission to pass | One to three failed mandatory lines is a resubmission. More than three is a fail |
| Flexible | The line is judged, and a failure is written up as a note rather than as a result | Never on its own a reason to refuse certification |
A failed mandatory line is never acceptable, and no result is ever reported as "passed with N of M". A fraction is how a mandatory failure gets rounded into a pass, so the report does not have the shape that would let it.
Who marks. One house reviewer, against this document, with Craig on every fail and on every locked-line failure. That is settled as D8 of the program and it is not delegable.
2. The Certified Implementer rubric
Marked on a real canvas, built from a blank business against a written brief. Not a fixture, not a demo account, not a canvas that already had seats on it.
| # | Line | Must |
|---|---|---|
| 1 | The business type is named, and the opening set placed is the one that type names | Mandatory |
| 2 | Every department that the brief's business actually runs carries at least one seat, or the canvas says in writing why it does not | Mandatory |
| 3 | Every seat placed comes from the catalog at the account's own tier | Locked |
| 4 | No recipe is placed that the account's tier does not reach | Locked |
| 5 | Every recipe carrying a gate has that gate drawn, and the person who passes it is named | Locked |
| 6 | Nothing that needs a person is set to run unattended | Locked |
| 7 | No client's personal data sits anywhere a published recipe would read it | Locked |
| 8 | Every port has something on the other end, or is listed as a deliberate open end with a reason | Mandatory |
| 9 | Every outgoing leg leaves the right side of its card; a bottom leg is a failure or a return | Mandatory |
| 10 | Payment arrives from the customer's right into Finance, and no line runs from Delivery to Finance | Mandatory |
| 11 | Every seat carries its agreement: what it takes, what it returns, the bar, and the cadence | Mandatory |
| 12 | The first run is executed on the business's own data, never on a fixture | Mandatory |
| 13 | The run's report is read back against the recipe's own checklist, and no mandatory line of that checklist is failed | Mandatory |
| 14 | The first month's report names what moved, what did not, and one thing to change, each with the number it rests on | Mandatory |
| 15 | Nothing the brief did not say is stated as though the brief said it; every gap is named as a gap | Mandatory |
| 16 | The units on both ends of every wire agree, or the disagreement is written down as a finding | Flexible |
| 17 | The canvas can be read by somebody who was not in the diagnostic | Flexible |
| 18 | The handover pack names who runs what, on which day | Flexible |
| 19 | The read-back uses the model's vocabulary rather than the candidate's own | Flexible |
Lines 3 to 7 are all of one shape: this candidate would do something to a client's business that we would have to undo. That is why they are locked rather than heavily weighted.
2.1 What the exam hands the candidate, and when
| Stage | What the candidate is given | What they hand back | Clock |
|---|---|---|---|
| 1 · The brief | A written brief for one business: type, size, what it sells, how work arrives, what it uses today, and three things the owner complains about | Nothing yet | Day 0 |
| 2 · The diagnostic | A 60-minute recorded session with a house reviewer playing the owner, answering only what the brief supports | A one-page read-back: what the business does, in the model's own words | Day 7 |
| 3 · The canvas | A blank business on their own Compose account | The canvas: seats laid out, house recipes placed, connections wired, gates drawn, every seat's agreement filled | Day 21 |
| 4 · The first run | The client data the brief supplies, as a file | One recipe run on that data, with its report, read back against the recipe's own checklist | Day 28 |
| 5 · The first month's report | A second data file representing month one | The report: what moved, what did not, and one thing to change, with the number each claim rests on | Day 35 |
The brief is deliberately incomplete in two places. Naming an absence is a pass. Filling it with a plausible invention fails line 15, which forbids exactly that.
2.2 Two stages the program has scheduled and this version does not yet mark
Both are designed and neither is live, so neither is marked today and a candidate is not examined on either. They are named here so that a partner reading this document knows what is coming and can see it arrive as a version change rather than as a surprise.
| Stage | What it will ask for | Where it was settled |
|---|---|---|
| A sixth stage: the business brain | A brain for the exam business with at least one confirmed claim in each of layers 1, 3, 4, 5 and 7, where the layer 7 claim could only have come from a seat holder rather than from the brief | The Business Brain Design §8.4 |
| A seventh stage, at day 21: the filled wrapper | Thirteen parts filled for the exam business, priced with its hours shown, marked on whether scope out is real, whether the promise has a measure the seat's own report can produce, and whether the tier named is one the recipes actually need | The Certified Partner Program §5.10.3 |
3. The Certified Author rubric
Marked on one recipe, taken through the recipe pipeline to a graded pass.
| # | Line | Must |
|---|---|---|
| 1 | A research document exists, carrying its source table and a stated rule for which kind wins on disagreement | Mandatory |
| 2 | Every principle is distilled rather than copied, and names the source it came from | Locked |
| 3 | No source's own wording runs past fifteen words, and there is at most one quotation per section | Locked |
| 4 | No practitioner's name appears in the recipe's title without a trademark license in hand | Locked |
| 5 | recipe-reviewed is empty on submission, and stays empty until the house review fills it | Locked |
| 6 | No client data appears anywhere in the recipe, its research, its fixture or its runs | Locked |
| 7 | The validator is clean | Locked |
| 8 | The graded run passes every mandatory line of the recipe's own checklist | Locked |
| 9 | The front matter carries all six load-bearing fields, each answered rather than left blank | Mandatory |
| 10 | Every setting has a default and names the line of the report it feeds | Mandatory |
| 11 | Every checklist line is marked mandatory, flexible or locked, and every locked line carries the sentence a person is refused with | Mandatory |
| 12 | The body has the eleven sections, in order, under their own names | Mandatory |
| 13 | Two graded runs sit in the run document, and a third where run 2 forced a change | Mandatory |
| 14 | The derived verdict word and the run's own word agree, or the disagreement is written as the finding rather than reconciled | Mandatory |
| 15 | No report reduces a mandatory failure to a fraction | Mandatory |
| 16 | The figure exists, is authored as SVG, is rendered to PNG, and its text fits inside its boxes | Mandatory |
| 17 | Where this stops says plainly what the recipe is not | Mandatory |
| 18 | No line the candidate added carries an em dash, and the spelling is US throughout | Flexible |
| 19 | The version notes say what each version changed and which run forced it | Flexible |
3.1 What the exam hands back, and when
| Stage | What they hand back | Clock |
|---|---|---|
| 1 · The proposal | The method, the seat it sits on, the ports, and one paragraph on why the catalog needs it | Day 0 |
| 2 · The research document | The source table, the sections, the checklist with each line's source, the settings, and the figure | Day 14 |
| 3 · The skill | Front matter, settings, checklist, the eleven sections, the version notes | Day 21 |
| 4 · The runs | Two graded runs in the run document, a third where run 2 forced a change; the validator clean | Day 28 |
| 5 · The house review | Nothing. This stage is ours | Day 35 |
The candidate proposes their method from a short list of gaps the house publishes. They do not choose freely, because choosing the easy one is the first thing a candidate does and the last thing a standard should allow.
4. Pass, resubmit and fail
4.1 The Implementer
| Result | What it takes | What happens |
|---|---|---|
| Pass | Every locked and every mandatory line passed. Flexible lines may fail and are written up as notes | Certified. Listed within five working days. The badge carries the date and the version of this document |
| Resubmit | No locked line failed, and one to three mandatory lines failed | One resubmission, of the failed parts only, inside 30 days. No extra fee. Marked by the same reviewer |
| Fail | Any locked line failed, or more than three mandatory lines failed, or a second submission still fails | Re-sit no sooner than 90 days, with a new brief and the full fee. Written feedback naming every failed line |
4.2 The Author
| Result | What it takes | What happens |
|---|---|---|
| Pass | Every locked and mandatory line passed, and the house review agrees the recipe is worth placing | Certified. The recipe is published with recipe-author naming them. recipe-reviewed is filled by the house, never by the author |
| Resubmit | No locked line failed, one to three mandatory lines failed | One resubmission inside 30 days, with the runs re-run rather than re-reported |
| Fail | Any locked line failed, or the house review finds the method sound and the evidence absent | Re-sit at 90 days. Where the method was sound, the same method may be resubmitted |
A candidate whose recipe passes every line but whose method the house review judges not worth placing is a pass on the exam and a decline on publication. Those are two different decisions and this document keeps them apart on purpose: the exam grades the craft, the review grades the catalog.
5. What the badge cites, and what happens when a line changes
A badge names the date it was earned and the version of this document it was earned against. That is what makes a rubric change survivable: a partner certified against version 1.0 is not retroactively uncertified by version 1.1, and a client reading a badge can open the exact version their partner was held to rather than the one current today.
| When | What happens to a partner already certified |
|---|---|
| A flexible line is added or reworded | Nothing. The badge keeps its version, and the new line is read at the next renewal |
| A mandatory line is added | Nothing immediate. The line is read and acknowledged at renewal, and the renewal spot check is marked against the version then current |
| A locked line is added | The same, with one difference: the line is sent to every certified partner when it lands rather than waiting for their renewal date, because a locked line describes something we would have to undo |
| A line is removed | Nothing. A partner is never asked to re-earn a badge against a shorter list |
Renewal is where a partner meets the current version. It requires the subscription still held, one canvas or one published recipe from the last year submitted for a spot check against the version then current, and every line that changed in the year read and acknowledged.
Changelog
| Version | Date | Change |
|---|---|---|
| 1.0 | 2026-09-19 | Written, on the day Craig started the build. Both rubrics moved here in full from The Certified Partner Program §3.1 and §3.2, with the marking scheme, both exams' stages, pass, resubmit and fail for each path, and what a badge cites. Nothing about any line changed in the move: the nineteen Implementer lines and the nineteen Author lines are the same lines, with the same standing, that document carried at v1.6. New here, and not in that document: the statement of what happens to a partner already certified when a line is added, reworded or removed, and the note that the two further exam stages the program has scheduled are not marked in this version |