Assessments · Readiness
Where it will stop,
and why.
Eleven questions about a deployment you are planning or already running — and the stage at which comparable ones stop.
Run it if something works in testing and has not shipped.
Eleven questions on a deployment you are planning or running: whether the capability is proven, whether the evidence requirement is understood, who will accept the output, and whether the old process still runs. Returns the stage at which comparable deployments stop, what your answers imply about yours, and what would move it — with the reasoning shown.
What it is looking for
Almost none of them fail on accuracy.
Which is why asking about the model first is the wrong place to start.
The usual internal explanation for a stalled deployment is that the model was not good enough. It is easy to check: ask whether anybody disputed the outputs during testing. Usually nobody did.
What actually happens is narrower. At some point the output leaves the system and enters the operation, and somebody becomes accountable for a decision they did not personally reason through.
Nobody refuses — refusing starts a conversation. They simply do not get to it today. The deployment does not fail; it queues, and queuing looks like progress from a distance.
So this assessment spends most of its questions on the handover rather than on the technology. That is where comparable deployments stop, and it is the cheapest thing to fix because it is a decision rather than a build.
What this costs when it is missed: the pilot budget, the sponsor's credibility for the next one, and the year in which the operation carried on exactly as before.
Two ways to run it
Free, or read by a person.
Same eleven questions either way. What changes is who reads the answers.
Free
Result immediately
Automated
A rule set reads your answers and returns a written finding. Nobody looks at it, and the report says so on every page.
- Where the work stops, on five named stages
- Your own answers quoted back
- Next steps, in the order they should happen
- What the diagnostic cannot tell you
Five working days · includes a call
Reviewed
Somebody reads your answers against deployments in your sector specifically, rewrites the finding for your situation, and talks it through.
- The finding rewritten rather than generated
- Read against comparable operations in your sector
- A thirty-minute call on what it means for you
- The questions the form could not ask, asked
And a third thing, which is not an assessment. A diagnostic engagement does not read your answers at all — it reads your operation, through interviews and your own records. The free assessment responds to what you told us. The reviewed one responds to what we found.
What comes back
A stage, named — not a score.
On screen immediately, in full, before anything is asked of you.
Where the work stops
Your position on five stages — capability, evidence, acceptance, workflow, production — with the stage named rather than numbered.
Your own answers, quoted
Especially the two or three that produced the finding, so you can see the reasoning and disagree with it.
What each stage showed
Including the ones you have already passed, because knowing what is in hand is as useful as knowing what is not.
What to do next, in order
Sequenced, because doing these out of order is the common way the work becomes expensive.
Not: an assessment of your model
Nothing here measures your model or your data. Any tool claiming to from eleven answers is guessing, and the result says so.
Worth saying
The likeliest outcome is not failure. It is two quarters of promising.
Nothing breaks. Nobody is blamed. The deployment is described as promising while the operation continues exactly as it did before.
That is worth stating plainly, because from inside it does not feel like a problem — which is why it goes on so long.
Frequently asked
Answers before you start.
Can I run it before anything is built?
Yes, and that is the better time. Run before the build, the questions are a specification. Run after, they are a diagnosis — and the second is more expensive to act on.
Does it assess our model?
No. Nothing here measures accuracy or data quality, and the result says so. It assesses whether the operation around the model is ready for it.
What are the five stages?
Capability, evidence, acceptance, workflow, production. Most deployments reach the third and stop, which is why the assessment spends most of its questions there.
Do I have to give an email address?
Not to see the result. It appears on screen in full, and an email is only needed if you want a copy sent to you.
We are not in a regulated sector. Is it still relevant?
Usually. The pattern holds anywhere a wrong output has a named consequence. It matters less where nobody has to accept anything, and the result will say so.
What does the reviewed version add?
A person reading your answers against deployments in your sector rather than a general pattern, a finding rewritten for your situation, and a call. If the free result already named the stage, it adds nothing.
Find out which stage yours reaches.
No email needed to see the result. If it tells you the deployment is ready, that is the finding and it goes out unchanged.