game, game. repeat.

Ways to Double Productivity on an Engineering Team

Thu 26 February 2026 #leadership

Yes way

each way is another 2x

the ways multiply, so if you do all of them really well

that’s 2^9 = 512x

versus doing none of them at all = 1x

productivity = value creation, e.g. growth rate, mission progress, KPIs

Survivor bias

sounds ridiculous until you consider…

1% of companies make it to year 1

1% of THOSE make it to year 5

aka 1 in 10,000 make it to year 5

most companies fail

most companies that survive are probably somewhere above 1x

on the 1x to 512x journey

and Alex Honnold free soloed El Capitan

and Arnold Schwarzenegger exists

Assumptions

most of the team is a good culture and skill fit…

the teamwork and morale lines will be especially unfixable without that

Stuff I’ve Done

  1. teamwork: golf -> basketball, e.g. mobbing, scrum, The Advantage, Wooden on Leadership
  2. morale: low -> high, e.g. work is a slog vs I can’t wait to get into work on Monday because I’m doing the best, most interesting, most fun, most important work of my life
  3. product: gut -> data driven ROI feedback loop, e.g. A/B testing, outcome reviews, learning points
  4. DORA + devX: low -> high performance org, e.g. good automated test coverage, CI/CD automation, good docs, plenty of testing environments, budget for tools, budget for training (i.e. time and money)
  5. history: undocumented -> well documented and easy to use, e.g. ADRs, GDDs, A/B test results and analysis, everything easily searchable

Stuff I’m Trying

  1. skill w/AI: none -> expert
  2. code: AI/newb unoptimized -> optimized, i.e. a junior dev or AI can make improvements to production safely in any particular area of the codebase, with limited context, without talking to someone else
  3. roadmap: AI unoptimized -> optimized, i.e. the projects at the top of the priority list are easy for AI to help with, so their costs are lower (than the same project without AI), so their ROI is higher (than the same project without AI) and that’s how they got to the top of the priority list
  4. data: AI unoptimized -> optimized, i.e. all data cleaned up and available from one place, we can ask questions of product usage logging tools, internal tools, error logging tools, and others and get very useful answers very quickly