A harm score is proposed as a complement to RICE, ICE, WSJF, and MoSCoW prioritization frameworks, addressing their blind spot around severe, low-reach harms. The score rates severity, reversibility, vulnerability, and recoverability of a feature's worst realistic failure on a 1-5 scale, using the highest single factor as the overall score rather than an average. Real-world examples include Australia's Robodebt scheme and Chime's account freezes. A worked RICE example shows how an account-freezing feature scores high on harm and gets moved into a critical-harm lane, forcing its safeguard (human review and appeals) to be bundled with it rather than left at the bottom of the backlog.
Table of contents
Why product prioritization frameworks can miss user harmLow reach can hide severe consequencesOver 200k developers and product managers use LogRocket to create better digital experiencesBuild a harm score with four factorsModify, gate, or escalate based on harmHow harm scoring changes a RICE decisionAdd a harm review to your product backlogLogRocket generates product insights that lead to meaningful actionQuestions this post answers
What is a harm score in product prioritization and how do you calculate it?
A harm score rates a feature's worst realistic failure across four factors: severity, reversibility, vulnerability, and recoverability, each scored one to five. The overall harm score is the highest of the four ratings, not an average, so a critical severity rating cannot be diluted by lower scores elsewhere. It is used as a decision trigger, not a precise harm prediction. See how other teams are combining risk scoring with RICE and similar frameworks on daily.dev.
Why does RICE prioritization fail to catch severe harm affecting a small group of users?
RICE multiplies reach, impact, and confidence, then divides by effort, so a feature helping ninety percent of users while seriously harming two percent earns most of its score from the majority. Safeguards built for the smaller, more vulnerable group score low on reach and fall to the bottom of the backlog, even when the harm to that group is severe and hard to reverse. Product managers weighing RICE against risk-aware frameworks can follow this debate on daily.dev.
What happened in Australia's Robodebt welfare scheme?
Robodebt used annual tax data to estimate welfare recipients' bi-monthly income and raised debts against people based on those estimates, without accounting for the difficulty of challenging a false debt. The 2023 Royal Commission into the Robodebt Scheme called it a crude and cruel mechanism that was neither fair nor legal, and it affected hundreds of thousands of people, causing severe financial and psychological distress. Anyone designing automated decision systems can track lessons like this via daily.dev.