Rewards go to dramatic firefighting while prevention goes unseen, so the people who cause or worsen crises keep getting celebrated as the ones who saved the day.
An example
The promotion committee at a payments software company had two site reliability engineers from the same platform team up for staff level. One had led five high-visibility incident recoveries that year, each written up in the company newsletter. The other had no incidents to her name, because her work on capacity alerts had removed two failure modes before they ever fired. The engineering director preparing the cases noticed that three of the first engineer's five recoveries had followed changes he had shipped himself, and that none of the post-incident reviews had asked what started the fire. The committee had already read the newsletter. The director added a page to both packets: incidents caused, incidents resolved, and failure modes removed, with platform uptime before and after each engineer's work. The committee promoted the second engineer and asked for the same page in every infrastructure case.
Why it works
Status, spotlight, and promotion attach to the save, so firefighting looks better, and fires look cheaper to start because the people who start them collect the credit for ending them. Planning looks worse, since its output is a non-event no one sees, and costlier, because planners must argue for resources against crises that have not happened and take the blame when one comes anyway. It thrives where performance is judged by visible events, post-incident review is shallow, rewards flow through storytelling, and the same person can plausibly cause and resolve a crisis.
The thinking behind it
Jonathan Baron Baron and Hershey's Outcome Bias in Decision Evaluation (1988) found that the same decision is rated as better made, and its maker as more competent, when the outcome happens to be good. A recovered outage is a good outcome, so the engineer at its center looks competent even when their change caused it. Post-incident reviews that trace causes put the original decision back under evaluation.
Daniel Kahneman Tversky and Kahneman's Judgment under Uncertainty: Heuristics and Biases (1974) describes availability: judging how often things happen by the ease of recalling examples. Dramatic recoveries written up in a newsletter are easy to recall, and failures that never happened leave nothing to recall. A page that counts incidents caused and failure modes removed gives the promotion committee the missing instances in a form it can see.
Warning signs
Incidents recur in one domain, and the same names appear as both cause and resolver. Post-incident reviews credit the resolution without tracing the cause. Promotions follow the rescues. People whose work prevented failures get no credit, and some of them leave. Leadership talks about heroics in town halls and rarely about uptime or other steady-state measures.
How to defend yourself
If you are the planner, make the absence visible: record the failure modes you removed, what they would have cost, and the steady-state measures that improved. Leave the firefighter's skill alone, since genuine emergency response is worth having. Argue for counting both sides of the ledger.
- Ask for post-incident reviews to trace each incident to its cause, including who shipped the change.
- In promotion cases, ask for incidents caused and failure modes removed alongside incidents resolved.
- Find an ally who controls measurement, such as finance or a risk function, to report steady-state performance.
How a leader shuts it down
Measure steady-state performance alongside saves, and require post-incident reviews that trace causes. Give prevention the same weight as rescue in promotion cases, and record contributions in a ledger as well as in stories. Keep rewarding genuine emergency response, and make the people who start fires visible.