We drive for the conditions

Decision making in software is similar

Drive for the conditions

Speed by default

Slow for hazards

When to slow down?

Exercise

Choose a platter and eat what's on it

Algorithm 1: Look Inside

Algorithm 1: Look Inside

Algorithm 1: Look Inside

Algorithm 1: Look Inside

Challenge

What if it takes some effort to investigate each option?

Algorithm 2

Explore Everything

• Easy to go through a bead door
• Explore until we find good food

Challenge

What if exploring one prevents exploring another?

Algorithm 3

Research

• Smell when the doors open
• Ask a passerby to check for you
• Send one person of your group through

Explore everything

Works well when it's cheap to experiment

Research

Works well when experimenting is expensive

The Planning Tradeoff

The higher the cost to experiment, the bigger the budget for research and planning

Bead Door Examples

• Choosing a library to transform data
• Class structure and naming
• UI layout

One-way Door Examples

• Releasing a public API to customers
• Adopting a library's data format throughout our code
• Moving customer data into a new database
• Choosing a translation strategy and using it throughout the UI

The Planning Tradeoff

The higher the cost to experiment, the bigger the budget for research and planning

Hack: make 1-way-doors 2-way

• Pre-release the API to one customer with an agreement that it might change
• Reduce the scope of the experiment
• Make it easy to reverse

Examples: de-risking decisions

• Wrap the library to convert the library data format to one we choose; couple to our format
  • (This is an anti-corruption layer)
• Prove the database with sample data (e.g. dual write)
• Prove the translation strategy with a small part of the UI
  • (i.e. Replace the decision with an experiment)

Alex's suggestion

1. Default to fast: assume bead door
2. Assess how reversible is the decision you're about to make?
   • And act accordingly

Alex's suggestion

When you identify a hard-to-reverse decision

In this case, "try the first idea" won't make you faster

1. If possible, reframe the decision as a reversible bead door decision
2. Otherwise, take time to plan and research before acting

Things to say in the mob

Things to say in the mob

• probing: "If we do X, can we change our mind?"
• gas pedal: "This looks reversible, can we try something?"
• pump the breaks: "This seems like a significant decision, can we write a plan?" (Consider writing an ADR.)
• pump the breaks: "Before we commit, could we run a reversible experiment?"

Speed by default

Slow for hazards