Personal effectiveness · Field guide
Built for Bad Days
The engineers who keep banking apps and airline booking systems running start from one assumption: some part will fail today. They do not try to stop every failure. They plan how the system behaves while a part is down and how fast the system gets back. This page draws twelve of their patterns as small pictures, each with a scene from ordinary office work, because the same patterns help people and teams get through a bad week.
Guess first
Same outage, different week
Two claims teams in the same company lose the claims system for a full Monday. Both teams have eight people and the same backlog of 400 open claims. Team A is back to a normal backlog by Wednesday. Team B is still catching up the following Tuesday. Which difference most likely explains the gap?
Tap an answer
The model
Measure the way back, not the hit
Resilience is how fast you return to normal after a setback. The size of the setback matters less, because you rarely choose the size of the hit, but you can prepare the way back.
Engineering teams measure the same thing. In the research behind the book Accelerate, the time a team takes to restore service after a failure was one of four measures that separated strong software teams from weak ones (Forsgren, Humble and Kim, 2018).
Every pattern below does one of two jobs. A pattern either makes the hit smaller or makes the way back shorter. The patterns fall into three groups, by when you use them.
The scenes and numbers on this page are made up to illustrate each pattern, and none of the numbers describe a real company.
Start here · The alarm
Worry can be an alarm you cannot act on
The grid makes one narrow point. Worry adds a cost before the result is known, and when you cannot change the result, the worry does not change the result either. Whichever way the result goes, the row where you worry carries one more unit of suffering than the row below.
Engineers who run systems around the clock learned the same lesson about alarms. A warning that wakes someone at 3 a.m. for a problem that person cannot fix trains the team to ignore warnings. Google’s guide to running large systems sets the rule that every alarm that calls a person must be something the person can act on (Beyer and others, Site Reliability Engineering, 2016).
The rule does not say “stop caring”. The rule says to sort each worry into the ones you can act on and the ones you cannot, and to spend your attention on the first group.
Group 1 · Before it breaks
Decide on a calm day
The decisions that matter most in a bad week are cheapest to make in a good one. These four patterns are choices you make in advance, so that when something fails, nobody has to think from scratch while under pressure.
Group 2 · While it breaks
Bend so you do not snap
When a part has failed, the instinct is to push harder at the part that failed. These four patterns do the opposite: they stop the push, slow the push down, keep the important work running or stop the damage from spreading.
Group 3 · After it breaks
Get back, then get better
Recovery has two parts: returning to a state that works, and changing something so the same failure costs less next time. These three patterns cover both parts.
Your plan
Plan your next bad day on a good one
Pick one thing that could go wrong in the next three months, at work or at home. Answer five short questions, one for each of the patterns you can set up today. The card on the right fills in as you type, and you can print the card or copy the card as text. Your answers stay in this browser.
My plan for a bad day
Written on a good day,
Use it this week
Pick one pattern and set the pattern up on Friday
Choose the pattern from this page that would have helped most in your last bad week. On Friday afternoon, spend fifteen minutes setting that one pattern up: write down the time limit, name the spare or agree the reduced mode with your team. A pattern set up on a calm Friday is ready when Monday goes wrong.
You rarely choose how hard the hit is, but you can choose how short the way back is.
Recovering well from setbacks is one part of digital intelligence, and the 30-day course builds the rest of that skill one day at a time.
Sources
- The worry grid comes from a reel by @damianiszn on Instagram (November 2026), redrawn here in our own colours. The saying behind the grid, that worrying about something makes you suffer twice, is older, and we could not trace who said the saying first.
- Nicole Forsgren, Jez Humble and Gene Kim, Accelerate: The Science of Lean Software and DevOps, IT Revolution, 2018. Time to restore service is one of the four measures of software delivery the book reports.
- Betsy Beyer, Chris Jones, Jennifer Petoff and Niall Richard Murphy (eds.), Site Reliability Engineering: How Google Runs Production Systems, O’Reilly, 2016, chapter 6, “Monitoring Distributed Systems”: every alarm that calls a person should be something the person can act on.
- Michael T. Nygard, Release It!, Pragmatic Bookshelf, 2007. Source of the names circuit breaker and bulkhead for software, and of the advice to put a time limit on every call to another system. Martin Fowler’s article “CircuitBreaker” (2014) explains the circuit breaker in plain terms.
- Marc Brooker, “Exponential Backoff and Jitter”, AWS Architecture Blog, 2015. Source of the advice to wait longer after each failed try and to add a little randomness to the wait.
- Yury Izrailevsky and Ariel Tseitlin, “The Netflix Simian Army”, Netflix Technology Blog, 2011. Netflix described switching off its own servers on purpose to practise recovering, which later grew into the field called chaos engineering.
- John Allspaw, “Blameless PostMortems and a Just Culture”, Code as Craft (Etsy), 2012. Source of the blameless review after a failure.
- Single point of failure, graceful degradation, blast radius, rollback and safe to repeat (engineers say idempotent) are common engineering terms with no single author. The explanations, the scenes and the three groups on this page are our own.