[news_to_me]

Also known as “PDSA” in some school districts that tried to implement it some years ago. It was one of the fad management techniques school districts make their faculty do, and predictably, wasn’t very effective.

Presumably the original PDCA has some good if applied correctly - I haven’t studied it, but have heard good things.

[ziggus]

Yeah, it's only applicable to processes that are repeatable and have some amount of measurability - hence the natural application to manufacturing.

I'm not surprised that managers in non-manufacturing/technical areas picked up on it, since it's easy to understand at a glance. However, it's difficult to implement well.

[marcosdumay]

If you know of any person that doesn't deal with repeatable processes all the day, I would really like to hear an example. Specifically on computing (most people here are software developers, aren't they?) the entire attention is focused on creating repeatable processes, that then we use computers to repeat for us.

PDCA means basically that you test a change before completely committing to it. That concept is the same as our canary releases, for example, but more clearly, it's the same idea as running one or two iterations of your algorithm on paper before you go and write the program.

It's not some niche thing that is rarely useful, it's just that people try to push it into exactly the work that can't use it.

[throwaway2331]

> any person that doesn't deal with repeatable processes all the day

Anyone that deals with people, no matter what all the training/literature/latest pop-industrial-psychology fad espews.

>PDCA:

Plan (see: "analyze data," i.e. read tea leaves (metrics)): impossible to do with people. Some try to get a "quick glance" via quantitative data, but that doesn't provide you anything practical, except ammo for political moves. People aren't processes. And any attempt to analyze them as a system is the peak of arrogance.

Do: i.e. either go with your gut, allowing your intuition to navigate you, or go with policy/procedure/a process to cover your ass, and check off the "I'm adding value" box, so you can spend your mental energy on other schemes.

Check: see: "Do." Either you know how a certain action/policy affected someone or you have to read "data" to figure it out. Average case, you redefine your metrics and massage the data to get the result you wanted, so...

Act: you can use your results as ammo for political moves. Pat yourself on the back, have your assistant make a spreadsheet, and then hold a 30 minute jerk-off session/meeting with your superiors and fellow sycophants.

Now that I'm really looking at it, this is already implemented by those who deal with people.

[robga]

In these cases you could use Boyd's OODA loop. It can be positioned as more able to deal with less certainty, epecially those that need to understand external context. I see echoes of it in Agile.