3 builders who never spoke to each other

Price the months a rebuild will actually cost

Infrastructure work is paid for in the switching features you did not ship during it, so count that before starting. Match the size of the decision to the size of the thing: if two people could build it in a week, skip the debate and reserve real research for choices that take months to reveal themselves.

The method

1. 01

   Count the cost before a rebuild: the months spent on infrastructure are months the switching features that win customers are not shipping.

2. 02

   Weigh the cost of deciding against the cost of doing: if two people could just build it in a week, skip the debate, and reserve the research for choices that take months to show results.

3. 03

   When a new requirement arrives, change the core of the system rather than bolting on another step or integration, accepting planned downtime now to avoid unplanned downtime later.

What it returned

• Fathom's Google Analytics importer was still unfinished months after it should have shipped because they chose infrastructure over features, and Jack Ellis admits that decision hurt them for a while even though revenue kept rising every month.

• He notes a few debates plus a time-boxed research project can consume a whole team for a week on something an engineer pair would finish in the same time.

Sources

• When should a decision be fast, or slow?Jason Cohen · longform.asmartbear.com · When should a decision be fast, or slow? · 2018-09-05
• We rebuilt our entire application - Fathom AnalyticsJack Ellis · usefathom.com · We rebuilt our entire application · 2022-10-11
• Simple Systems Have Less DowntimeGreg Kogan · gkogan.co · Principles for Simple Systems · 2020-03-02