What does deployed mean?
A tool is deployed when the process it touches produces to standard, at rate, at yield, without heroics, and stays that way. Five conditions, all of them about output rather than use.
Five conditions, all about output
To standard, so the result meets specification rather than approximately resembling it. At rate, so it holds under production volume rather than on a good afternoon. At yield, so the cost of the failures sits inside the economics. Without heroics, so it survives the absence of the person who set it up. And stays that way, because processes drift, tooling wears, suppliers change their own processes without mentioning it, and models degrade against inputs that have moved.
A system that met the first four on the day of handover and nowhere since was not deployed. It was demonstrated.
Which is why asset care, calibration, requalification and monitoring are not what happens after deployment. They are what makes the word mean anything past the first month.
Four ways it goes
A new inspection method reads to standard and keeps line rate, and nine months on the fixture has worn and it is passing parts it should reject: four conditions held, the fifth was never resourced.
A cell runs beautifully until the engineer who commissioned it moves to another plant, and stops making rate within a fortnight, because what was holding it together was a person adjusting it daily. It was never deployed. It was attended.
A qualified process running for two years starts producing scrap with nothing changed in the plant, because the supplier moved to a different feedstock and did not consider it a change worth notifying.
And a vision model qualified on last year's product mix reports high confidence on a variant it has never seen, which is the case where adoption looks best at the moment deployment is worst: the operators trust it completely, and the inputs moved.
It cannot be done from outside
This is where deployment parts company with adoption. The system pushes back while it runs, and nothing above it can respond at the speed that happens. The response comes from inside, from someone with the standing to change the process while it is running. That is horizontal agency in practice rather than a phase in a plan, and it is the Manufacturing Engineer’s position. See Systems Doing.
And it fails quietly
Adoption failures are visible: nobody uses the thing and everybody knows. Deployment failures happen while people are using the thing enthusiastically, and surface later as scrap, a rejected batch, an audit finding, a customer who stops calling. The lag between the failure and its discovery is the expensive part.
A programme that reports adoption is reporting attendance. Deployment is the only one of the two with evidence attached, and because the fifth condition never closes, that evidence has to expire. Ask for the yield number, ask when it was last requalified, and ask who on site can change the process while it is running. See Deployment Readiness for how the distance to it is produced, and Adoption for the necessary half this entry is not dismissing.
Questions
What does deployed mean?
A tool is deployed when the process it touches produces to standard, at rate, at yield, without heroics, and stays that way. Five conditions, all of them about output rather than use. A system that met the first four on the day of handover and nowhere since was not deployed, it was demonstrated.
What is the difference between deployment and deployment readiness?
Deployment is a state, deployment readiness is the measure of distance to it. Readiness tells you how close a capability is to producing its intended outcome reliably and at viable cost. Deployment is the point at which it does, and keeps doing it.
Why can deployment not be delivered from outside?
Because the system pushes back while it runs, and nothing above it can respond at the speed that happens. The response comes from inside, from someone with the standing to change the process while it is running. That is horizontal agency in practice, not a phase in a plan.
Why do deployment failures take so long to surface?
Because they happen while people are using the thing enthusiastically. An adoption failure is visible immediately, since nobody uses the tool and everybody knows. A deployment failure shows up later as scrap, a rejected batch, an audit finding, or a customer who stops calling. The lag is the expensive part.
Is deployment ever finished?
No, because the fifth condition never closes. Processes drift, tooling wears, suppliers change their own processes without mentioning it, and models degrade against inputs that have moved. Asset care, calibration, requalification and monitoring are not what happens after deployment; they are what keeps the word meaning anything.