Overproduction is waste, and you're probably doing it.

Happy Friday!

In the mix of all the AI conversation, there is a common thread about how much output it can produce. Other threads are about the quality of it and its consequences. In software, this has led to a lot of conversations about whether certain parts of our industry need to adapt now that AI is here. I already wrote a bit about how code review is finally getting scrutiny.

What nobody is talking about, though, is overproduction. Overproduction is waste.

When you produce more of anything than there is demand for, you have surplus. That surplus has to go somewhere. Physical goods have to move to a storage facility. This means you need not only the facility, but also transportation, security, and additional inventory management. That is a whole lot of cost to absorb for producing what nobody wants.

Software has always suffered because it's invisible and intangible. Software overproduction has the same problems. The waste is real, but we don't feel it the same.

In software, when we overproduce, we are producing change faster than the market wants it. We ship features, products, and fixes when there isn't demand. The interesting bit is that often this software just goes out to the users who never want it.

Know what happens when you give someone something they didn't want? They ignore it.

Organizations ship features nobody wanted as fast as they can, and those features never get used. This is different than doing something iterative- trials, pilots, and market tests because each of these terminates when there is no demand. If you're managing releases because your customers can't handle the amount of change, you're overproducing.

What happens when you keep putting in changes, features, and fixes nobody wants? The product becomes too complicated. You begin to lose customers. The glowing reviews of the past shift to resignation that they're too invested.

But wait. There's more.

Remember when I mentioned the part where you have to manage the surplus? That happens in software too! It shows up in increased development costs, as navigating all the code nobody runs is now part of the change sets, debugging, and fixes. In the warehouse example, it's like trying to find a pallet of goods in an unmarked maze.

Want to know the biggest tell of whether you're overproducing? When your releases and deployments are not aligned, or you use feature flags to control when people access things. These all point to the highly likely scenario that you're producing more than the demand.

This type of waste is insidious because it feels so good to do. We take pride in how quickly we can produce what no one wants. We revel in the extra costs of managing and working with all that code we built.

And all of this happened before AI!

I want to try to make what I write have something useful in it. So here are the two bits. First, I wrote an article about this topic that you can read more about. Second, some of you are wondering what on earth I mean by demand, because it isn't what comes from your boss or stakeholders. Demand is from the market.

The good news here is that knowing your demand quantitatively is a well-known metric called "Takt time." In other words, you can measure overproduction, see its additional costs, and then slowly eliminate it. Teaching leadership teams to build signals like this is exactly what my Measures Metrics and Signals Workshop does.

Let that sink in. You could eliminate waste and make your organization more effective and efficient by eliminating overproduction. Your outcomes would improve, your quality would improve, and your costs would drop.

AI is a ticket to overproduction. It creates waste, adds cost, adds to everyone's burden, and we all feel good about it.

What do you think? Reply and tell me. And if "we're getting behind, we need to improve velocity" is the soundtrack at your company, grab some time with me and we'll find out whether you're actually overproducing.

Sincerely,

Ryan