[jackero]

Maximizing productivity comes from maximizing efficiency. Efficiency is about sitting down and analyzing your tasks holistically and doing min-maxing to ensure every process achieves its greatest result.

Now ask yourself this: how often do people do this at all? Pretty much never. Most of us only do it when you have to, because you aren’t making enough money, because your application is slow, because you can barely meet budget, or because you are trying to land on the moon and failing costs too much.

[jgeada]

The other hard lesson learned during the Covid shutdowns (actually, it is learned and promptly forgotten at every major crisis):

maximizing efficiency also maximizes the number of single points of failure in your system. Anything that goes wrong breaks an optimized for efficiency system.

You need to have resilience and redundancy to deal with variability, but those cost money.

Much better from the management perspective to ignore those risks, cash in on the cheap profits and blame "unexpected events" and get a bailout when things go wrong. They cash in on the easy money and have no downside consequences.

[TheOtherHobbes]

The article actually says this.

I'm wondering how many people read it to the bottom? It makes the case that productivity analysis tools already exist, maximally productive systems have some slack in them and 100% efficiency isn't just impossible, it's counterproductive (for the reason you state).

But the problem with AI is that it adds a random element which makes any kind of modelling much harder. Sometimes you get good results fasts, sometimes it wastes a lot of time and holds up the project.

You never know what you're going to get. So any kind of project planning becomes even harder.

But that's not even the main point. The subtext - which isn't stated - is that the C-suite has persuaded itself that AI is a system that is more controllable and predictable than human employees.

When in fact - as anyone working at the coal face knows - it's the opposite.

And that's a problem, for all kinds of reasons. The obvious ones, like the loss of expertise through career progression, have already been talked about.

The less obvious one being discussed here is that the more AI is used, the less predictable all kinds of projects become - both in time and quality.

And if the economy is now being designed on the assumption the opposite is true, that's not going to end well.

In previous phases of industrial revolution consistency was the bedrock benefit.

Trying to create a revolution out of inconsistency is a very risky click.

[danaris]

> You need to have resilience and redundancy to deal with variability, but those cost money.

Reiterating for truth, and also to expand upon the point:

These are things that cost money all the time, but only pay off visibly in a crisis. And it has to be a crisis of the right kind (if your headquarters burns down in a wildfire, it won't help you to have 225% coverage on every role). So this makes it very difficult to justify to people who think only numbers—and only specific kinds of numbers—matter.

But redundancy and resiliency, at least in most cases, also make the lives of everyone working there better. They mean, among other things, that if one person needs to get surgery, or take their child to the doctor, or just go on vacation, there's still enough people there to keep the work flowing smoothly. The people still there won't be hopelessly overloaded, the work will get done, and the one person who's out won't have a mountain of catch-up work to go when they get back.

The only drawback is that it means you're paying people to work at a rate that means they regularly have downtime and aren't "fully utilized" constantly. (By definition.)

[hinkley]

Onerous tasks can also be shared between redundant positions so that o one person has to do them all the time. I’ve left places that were spiraling specifically to avoid being the last person who knows a terrible tasks. Sorry other guy.

[toast0]

> The only drawback

You also spend a lot more time on communication. If you have one person who is the resource for X, they can spend their time doing X and don't have to spend time on coordination. When the procedures for doing X change, only one person needs to figure it out, etc.

That doesn't mean having a single person is the right decision, the benefits of having multiple people are important; just want to be clear about the drawbacks of multiple people doing the work.

[hinkley]

The lemonade to be made however is that if you don’t talk about your work you don’t reflect on it, and it’s more difficult to improve if you don’t examine your work and the work of others performing the same task.

That does make you more disposable, but also more useful if you can embrace it.

[danaris]

...But, on the flip side, that extra communication, and especially making sure that procedures are documented somewhere so that everyone can reference it, rather than just having it all live in one person's head, are vital for institutional stability and continuity.

[hinkley]

Queuing theory. You can only fill a pipeline around 60% full before you start seeing measurable delays, and 80% is where the wheels begin to come off. The line goes very vertical shortly after 80%.

[boothby]

> Efficiency is about sitting down and analyzing your tasks holistically and doing min-maxing...

Accurate analysis depends on accurate data. Which is why some people I know are required to account for their work activities in 7-minute increments, have frequent and detailed meetings to account for progress, project planning, etc.

Actually, doing that analysis has a cost that must be discounted from any perceived gains of efficiency, to be truly efficient. One learns to tolerate some waste to avoid the ultimate time-suck.

To say nothing of maintence and sustainability -- if you're always sprinting, you're doomed to fall.

[Sevii]

For factory optimization they didn't ask people to track their hours at 7 minute intervals. They put a camera on them and analyzed what they did over an entire shift. Then they redesign the factory line to optimize things. Having people manually track time has far more overhead and is less accurate in general. In practice I've never seen a software org take estimates or time tracking seriously beyond "Track your hours in 15 min increments against tickets".

[hinkley]

I will have this written on my gravestone if Edward Tufte doesn’t beat me to it:

Charts are for asking questions, not answering them.

That analysis phase can be done brilliantly, or disastrously. It’s down to whose hand is on the rudder.

[7952]

7.5 would be nicer as it fits into sixty minutes!

[hinkley]

Boeing does 6 minutes, which means that it’s impossible to have a 45 minute meeting. My coworkers decided that if the meeting went over its .8 hours and if we cleared out on time or earlier that it’s .7 hours.

I don’t know why you would do 7. That doesn’t divide into 60 at all. Nor 480.

[hinkley]

Maximizing efficiency is how smaller companies eat your lunch. They maximize effectiveness instead. But that’s hard to measure except in customer satisfaction and market share. So it takes a lot of storytelling to get that far, and it can be broken quickly by the wrong execs being hired or the right ones leaving.

There’s a reason “bean counter” is an epithet.

[saulpw]

I disagree with the toplevel statement. In high-school science terms, productivity is like accuracy whereas efficiency is like precision. You can be very "productive" at a high-level without being efficient. Similarly you can be quite efficient at your tasks while not managing to achieve the toplevel goal.

You may argue that "analyzing your tasks holistically" will make you more efficient at achieving your goals. But min-maxing is for tactics, and is antithetical to strategy. Very often, taking a nap or a bath or walk will yield an insight that changes the entire game, obviating weeks or months of efficient and diligent work. We've all experienced this, so why do we insist on "maximizing efficiency"?

[hinkley]

The word “effectiveness” has a lot of utility in this sort of conversation.

It’s no good driving as fast as possible straight at a brick wall L.

[MaKey]

> Now ask yourself this: how often do people do this at all? Pretty much never. Most of us only do it when you have to, because you aren’t making enough money, because your application is slow, because you can barely meet budget, or because you are trying to land on the moon and failing costs too much.

I do it most of the time at work because I find joy in finding the best possible solution.

[marcosdumay]

The article is about several very widely known effects that don't require holistically analyzing anything nor come close to maximizing anything.