We had six features on the table and one sprint to fill. We had been arguing about the order for three weeks, which if you do the math is longer than the sprint itself. Sales wanted the Salesforce integration because one enterprise prospect had name-dropped it. Engineering wanted to pay down an API refactor that was making everything slower. Customer success wanted onboarding tooltips. I had been staring at a prioritization spreadsheet for two weeks, dragging rows up and down, trying to make everyone's favorite thing land at the top, which is a great way to make no decision at all.
The twenty-minute session that broke the tie
I wrote a one-paragraph description for each of the six features, being deliberately honest in the descriptions about how much evidence we actually had. For the Salesforce one I literally wrote 'one prospect asked, no other signal.' I pasted all six into the backlog_items field of the Score and Rank a Feature Backlog with RICE prompt, set the team size to five engineers, and ran it on Claude Sonnet 4.6, which I trust for this because the scoring requires nuance and the willingness to call a popular feature a weak bet.
The Salesforce integration scored 8.4: high impact, but the model dropped Confidence to about 30 percent because my own description admitted there was one data point, and Effort was medium-high. The onboarding tooltips scored 24.1: medium impact, but Very High reach because every single new user hits onboarding, and tiny effort. The numbers said the boring tooltips should go first. The numbers were right.
What a shared number did to the room
Here is the real value, and it is not the math. The math is simple; I could do RICE on a napkin. The value is that the number is external. When I say 'I think onboarding matters more,' that is my opinion against the sales rep's opinion, and opinions do not resolve. When the readout says 'tooltips: 24.1, Salesforce: 8.4, and here is the driver of each,' the conversation shifts from whose gut is bigger to whether the inputs are right. That is a conversation a team can actually have.
And it is a conversation that does not feel personal, which is the part I underestimated. The reason roadmap arguments get so ugly is that they become proxy wars about whose judgment the company trusts. The sales rep is not really arguing for Salesforce; they are arguing that their read on the market should count. When you put a shared scoring framework in the middle, you give everyone a face-saving way to update their position. Nobody has to admit they were wrong; they just have to agree the inputs were off. I have watched grown adults gracefully reverse a three-week stance in one meeting because the number gave them permission to.
- The sales rep's first question was 'what does Confidence mean,' which kicked off a genuinely useful discussion about the difference between one loud prospect and a validated need.
- The stress-test section flagged the API refactor's effort as likely underestimated, which opened an honest re-scoping instead of a optimistic guess that would have blown up mid-sprint.
- We shipped the tooltips first, measured a 22 percent drop in onboarding support tickets, and used that hard number to raise our confidence on the next batch of bets.
- Two of the six landed within 10 percent of each other and the prompt explicitly called it a toss-up and suggested 'which one unblocks a future feature' as the tiebreaker, which is exactly how we broke it.
Why this is the one I would pay for
Most of my prompts save me time. This one saves me from a bad decision, and a bad sprint costs five engineers a week, which is real money. The honesty is what makes it worth it: it is built to refuse to let a feature's popularity inflate its confidence score, which is the single most common way human-run RICE sessions lie to themselves. I have watched a room talk itself into 90 percent confidence on a feature one person wanted. This prompt will not do that, and that resistance is the product.
RICE is not magic and the scores are only as good as your inputs, so be brutally honest in your descriptions. But a structured, external number turns 'I think' into 'our stated assumptions say,' and that shift is worth twenty minutes. Grab the Score and Rank a Feature Backlog with RICE prompt on Prompt Dock and run it on your next contested sprint. I feed the winners straight into my Turn a Rough Idea or Slack Thread into a Team-Ready PRD prompt and go from ranked list to scoped work in an afternoon.