People, process, delivery and AI. Fixed properly, and measured.

Playbooks / Reliability · 2 min read

Stability without heroes

A hero is a single point of failure with a pulse.

The problem

“The same person fixes every incident”

In the model

Product and Operations · Reliability

In 30 seconds

  • Hero culture gets rewarded because heroics are visible and prevention isn't.
  • Map who can fix each critical system at 2am. One name means a risk.
  • Fix it by pairing on real work, not by writing more docs.
CRITICAL SYSTEM CAN FIX IT ALONE AT 2AM Payments BillingSingle point of failure Search Deploy pipelineSingle point of failure Auth
A bus-factor audit. Every system with a single name is a priority, whatever its tech-debt score.

The situation

Every serious incident ends the same way: the same engineer logs in at 2am and fixes it. Everyone is grateful. Nobody else learns how.

What's really going on

The hero gets thanked at the all-hands. The person who made the incident impossible never does. So the team learns that rescuing gets noticed and preventing doesn't.

What to do

  1. Run the audit. List your ten most critical systems. Write who could fix each one alone at 2am.
  2. Pair on single names. The expert drives the next change, someone watches, then they swap.
  3. Make on-call fair. A published rota, one week in four at most, and time off after a rough night.
  4. Ask a better postmortem question. "What made this easy to get wrong?" not "Who did this?"
  5. One owner, one action, one date. A postmortem with twelve actions has none.

How you'll know it's working

  • Repeats: the same incident type recurring less each quarter.
  • Spread: pages and call-outs shared evenly across the rota.
  • Care check: nobody on call more than one week in four.

In the frameworks

  • SAFe: Built-in quality: quality built in from the start, not rescued at the end.
  • DORA: Time to restore and change failure rate measure stability. A generative culture predicts it.
  • Lean: Stop the line: anyone can stop the work when they see a problem, and is thanked for it.
  • Scrum: A real Definition of Done prevents the late-night rescues.

In other industries

  • Healthcare The one GP who always stays late to clear the list.
  • Accountancy The manager who works every January weekend to rescue the tax returns.
  • Manufacturing The engineer called in every night shift because only they know the old press.
  • Small business The owner who is the entire backup plan.

What to say

To your hero, privately:

"You've saved us more times than I can count. But I've made you a single point of failure, and that isn't fair on you. For the next month, fix nothing alone. I'd rather we're slower than you being the only one who knows."

Try this week

Do the 2am audit for your top five systems. Count the single names.

Think about it

What does your team celebrate: the fire that got put out, or the week with no fires?