Killing performance ratings
Adobe killed ratings in 2012. The press loved it. The actual replacement — and the failure modes that followed — never made the keynote.
Killing performance ratings is the HR equivalent of switching to a standing desk: everyone announces they did it, almost nobody talks about month nine. The honest version is messier and more interesting.
I've worked with or studied eight companies that went fully rating-less. Five made it stick. Three came back to a modified rating. Here's what each replaced ratings with — and what the post-mortems look like.
- Frequent (monthly) check-ins replace annual review
- Forward-looking goals, not backward-looking scores
- Compensation decisions made by managers + comp business partner with calibration
- Promotion decisions on separate (more rigorous) cycle
- Examples: Adobe, Deloitte, Microsoft (partial)
- Replace rating with a short written narrative + 2–3 evidence artifacts
- Calibration meeting uses narratives, not numbers
- Comp tied to band + narrative + market
- Promotion has explicit rubric, separate from comp
- Examples: Netflix, Stripe (modified), GitLab
- 'Just remove the rating, change nothing else.' Three companies tried this. All three came back to ratings within 18 months because comp decisions had no defensible logic.
- Replacing a 1–5 scale with a 1–3 scale and calling it rating-less. Employees see through it in one cycle.
- Going rating-less without investing in manager calibration training. The result: same biases, no audit trail.
- Eliminating ratings but keeping forced distribution. The most cynical of all — the rank is still there, just hidden.
1. They replaced the artifact, not the conversation
Calibration didn't disappear. It moved from a number on a spreadsheet to a narrative in a meeting. The hard work didn't go away — it became more honest.
2. They separated comp, promotion, and feedback
The original sin of the 1–5 rating is that it tried to do three jobs at once: tell you how you're doing, decide your raise, and decide your promotion. Rating-less companies that work do all three jobs separately, with different cadences and different evidence.
3. They trained managers in writing, not in talking
When the rating becomes a narrative, the bottleneck becomes a manager's ability to write a defensible paragraph. That's a skill most managers were never taught. The successful companies invested in it.
Daniel Kahneman called it the 'WYSIATI' effect — What You See Is All There Is. When a manager has to choose between a written narrative and a numeric rating, the brain reaches for the number, every time. Numbers feel objective, comparable, and defensible. They are none of those things in practice, but the cognitive shortcut is irresistible. This is why rating-less programs that didn't replace the artifact with something equally legible quietly drift back to numbers — managers reinvent informal 1-5 scales in private spreadsheets within 18 months.
Carol Dweck's growth-mindset research adds the other half of the picture: numeric ratings activate a fixed-mindset frame ('I am a 3') while narratives activate a growth frame ('here's what I demonstrated and what's next'). The companies that successfully went rating-less did so because their narratives were structured enough to be calibrated AND open-ended enough to preserve the growth frame. Most companies pick one or the other and fail at the rest.
Microsoft killed stack ranking in 2013. The press treated it as the headline. What didn't make the headline was the four-year investment in 'Connect' conversations, manager writing training, and a calibration process that uses narratives but still rolls up to a distribution at the comp-decision moment. The number didn't disappear from the system. It moved out of the employee's view and into the calibration room — which is where the meaningful version of it always lived anyway.
- Can your managers write a defensible 3-paragraph performance narrative? If not, train first, kill later.
- Have you separated the comp decision from the feedback conversation? They need different cadences.
- Do you have a written promotion rubric that doesn't require a rating as input?
- Will calibration still happen — just verbally — or are you actually ending it?
- What's your 18-month commitment to NOT reinstating a rating when the first awkward comp cycle happens?