The problem isn't the framework
You've got your prioritization framework. Maybe it's RICE. Maybe it's a custom value/effort matrix your team spent three weeks bikeshedding. Maybe you're doing weighted scoring with seven different criteria.
Doesn't matter. They all produce the same result: a ranked list that feels scientific but collapses the moment someone asks "but how do you know this will work?"
Because the actual problem isn't which framework you use. It's that you're running a precision calculation on vibes-based inputs.
Reach score of 5,000 users? Based on what—your rough guess from that sales kickoff six months ago? Impact rating of 8/10? Says who—the PM who's been pushing this feature since Q2? Confidence level: medium? What does that even mean when the last time anyone talked to a user about this was never?
You can't prioritize what you don't understand. And you don't understand it because all your evidence is scattered across seventeen different tools, remembered differently by everyone in the room.
What "scattered research" actually looks like
Last week, your team had a prioritization meeting. Someone mentioned that enterprise customers keep asking for SSO. You remember a support ticket about it. Your CSM thinks it came up in three renewal calls. Engineering says it was in that Slack thread from the offsite. Your designer has notes from a user interview... somewhere.
Nobody has the actual quotes. Nobody knows if it's three customers or thirty. Nobody can say whether they're talking about the same problem or three different ones. So you do what every team does: you guess. You assign it an impact score based on who spoke most confidently. You move on.
This is how features get built. Not because they're the right thing. Because someone had the most convincing vibes in the room.
The research exists. Someone did those customer calls. Someone logged those support tickets. Someone wrote down those interview notes. But it's in:
- Google Docs that nobody titled consistently
- Notion pages nested four levels deep
- Slack threads that are now unsearchable
- Someone's head (they're on vacation)
- Meeting recordings nobody will ever watch
- That Dovetail account three people have access to
Your prioritization framework can't fix this. It's a sorting mechanism. You can't sort evidence you can't find.
The feature factory isn't a process problem
People love diagnosing feature factory syndrome as a process issue. "We need to be more strategic." "We need to say no more." "We need stronger prioritization."
Missing the point.
You're a feature factory because you have no reliable way to know what actually matters. So you build what's loud. What the CEO mentioned. What the customer paying $500K just asked for. What feels urgent.
It's not weak discipline. It's weak inputs. When you can't quickly pull up:
- How many customers mentioned this problem
- What they actually said (not your summary—the words)
- What segment they're in
- Whether it correlates with churn, expansion, or nothing
- What you already tried before and what happened
...then every feature request becomes equally plausible. Everything could be important. So you build it all, or you build whatever has the best internal champion.
A PM at a B2B SaaS company told me they spent two weeks building a reporting feature because their biggest customer demanded it. When they finally shipped it, they discovered that six other customers had asked for the same thing eight months earlier, but those requests were buried in different tools and nobody connected them. They'd have prioritized it way earlier with actual evidence of demand. Instead, they built it reactively and late.
That's not a prioritization framework failure. That's a "we can't see our own research" failure.
What changes when your evidence is actually accessible
Here's what good prioritization actually looks like:
You're in that meeting. Someone suggests the SSO feature. Instead of gut-checking it, you pull up:
- 23 direct mentions across customer calls and support tickets
- 14 of them from enterprise segment
- 8 mentioned it as a blocker to expansion
- 3 mentioned it during churn calls
- First request was 11 months ago, most recent was yesterday
Now you can score it properly. The Reach isn't a guess—it's 23 confirmed data points, concentrated in your target segment. The Impact isn't a vibe—8 people literally said it's blocking money. The framework doesn't change. But suddenly it's doing what it's supposed to: helping you compare real evidence instead of competing opinions.
Mimir helps with exactly this—connecting customer feedback, interviews, and usage patterns so you can actually see what you know. Not as a replacement for thinking, but so your thinking is based on evidence instead of whoever remembered the thing most recently.
The frameworks aren't the problem. They're fine. RICE works. Value/effort works. Weighted scoring works. They all work if you can feed them real data.
The problem is you're trying to run quantitative prioritization on qualitative mush. And no framework can save you from that.
Stop optimizing the wrong thing
If you're arguing about whether to use RICE or ICE or some custom thing, you're optimizing the wrong layer. The framework is the easy part. The hard part is having your research organized well enough that scoring isn't just collaborative fiction.
Your prioritization is only as good as your ability to answer: "How do we know this matters?"
If the answer is "I'm pretty sure someone mentioned it," your framework won't help you.
Fix your inputs. The framework will take care of itself.