One-way-doors and two-way-doors

This is a technique I’ve learned three years ago at AWS. I still use it daily. Here’s Jeff Bezos personally explaining it.

The doors are a visual tool for a better perception of the implications of the decision we’re about to make. When presented with a decision, we map the solutions onto a door type and then evaluate the proposals.

The tool helps us make better decisions faster. It also helps us not make too many bad decisions. And the cool thing is that we can use it for any type of decision.

We’ll assume each proposal was made deliberately and with insight.

Two-way doors

If we proceed with this decision, we can always come back.

This is why we call it a two-way door. There’s a way through, and a way back. It may not always be easy, and we may have to deal with some consequences, but we can always ‘revert’.

Running a pilot deployment for one products to test out cloud vendor reliability is an example of two-way doors. So is feature flagging a new search algorithm and rolling it out to 10% of users. So is asking a senior IC to lead a project as a trial run before the formal promotion.

Each of these decisions is concrete and directed, but we’re always able to change our mind based on the new information. We can then make them more quickly and with less deliberation.

This isn’t an excuse to make uninformed decisions, rather a tool that helps us avoid ruminating over low-implication decisions.

One-way doors

If we proceed with this decision, we’re sort of locked-in.

This is why we call it a one-way door. There’s a way through, but the way back is long, hard or even impossible. If we’re headed for these doors, we better think it through. And then some more.

Signing a 3-year enterprise contract with AWS, locking into cloud-first strategy is an example of one-way doors. So is doing a product-wide search algorithm replacement with breaking changes and no rollback options. So is laying off your core infrastructure team due to AI advancements.

No one is saying these decisions are necessarily bad or that they won’t pan out well. They might. But if they don’t, we have a problem, because we’re so deeply invested in them.

Quiz time

You’ll be presented with decisions, and need to classify them as one-way or two-way doors. After you classify it, try to turn it into the other door type.

Hope you’ve had some fun and learned something new!