What Mimir thinks deserves your attention, with evidence and rationale for each.
Recommendations are Mimir's primary deliverable — an AI-generated, prioritized list of what deserves attention. Each recommendation is grounded in your insights and your business context, so they're specific to your product, your users, and your actual data.
These aren't generic best practices or industry benchmarks. When Mimir recommends “Add bulk export for enterprise accounts”, it's because multiple sources in your data describe the problem, the themes support it, and your business context (user segments, goals, constraints) makes it the right priority. Every recommendation comes with a rationale explaining the reasoning.
The recommendations list supports multiple ways to find what you're looking for.
Sort options
Sort by quick wins (high impact + low effort), highest impact, lowest effort, status, or AI-recommended order. Quick wins is a good default for finding things you can ship fast with meaningful results. AI-recommended order reflects Mimir's overall priority ranking based on all available signals.
Filter by status
Focus on recommendations at a specific stage: New, Exploring, Building, Shipped, or Rejected. Useful when you want to see only what's in progress or review what's been decided against.
Filter by impact and effort
Narrow down by impact level (High, Medium, Low) or effort level. Combine with sorting for precise views — for example, filter to high impact + sort by lowest effort to surface the best opportunities.
Every recommendation has a status that tracks where it is in your decision-making process. Status persists across re-analyses, so your progress is never lost when new data comes in.
New
The default state for freshly generated recommendations. You haven't evaluated this one yet.
Exploring
You're actively evaluating this recommendation — gathering more context, discussing with the team, or refining the approach. A good signal to stakeholders that this is under consideration.
Building
You've committed to building this. The recommendation has moved from analysis to execution.
Shipped
Launched and live. Marking a recommendation as shipped helps you track what percentage of your data-driven insights are turning into real product changes.
Rejected
Decided against. This is just as valuable as shipping — it documents that you considered the recommendation and chose not to pursue it. Rejected recs are filtered out of the default recommendations view.
Click any recommendation from the list to open its detail page. This is where the full picture comes together.
Rationale
The core of the detail page. A written explanation of why Mimir recommends this, what problem it solves, and how it connects to your data. The rationale is generated by Sonnet (Claude's strongest model) to ensure the reasoning is clear, nuanced, and specific to your situation.
Supporting evidence
Direct quotes and observations from your sources that support this recommendation. Same drill-down pattern as insight detail pages — every claim traces back to what real people said.
Linked insights
Which themes this recommendation addresses. Click any linked insight to jump to its detail page and see the full evidence. This connection is what makes recommendations evidence-based rather than opinion-based.
Recommendations aren't set in stone. When you're viewing a recommendation in the workspace panel, simply chat about it naturally — Mimir detects the context and can update the rationale in real time.
As you discuss changes with Mimir — adjusting scope, questioning assumptions, shifting priorities — the rationale text on the detail panel updates live. You see the changes streaming in as the conversation progresses.
This is one of the most powerful features for product managers who need to shape recommendations before sharing them with a team. No special commands required — just open a recommendation and start talking.
Use status to communicate with your team
Status isn't just for your own tracking. When your team can filter by Exploring, Building, and Shipped, everyone has the same view of what's being considered and what's in progress. It replaces the “what are we working on?” thread.
Refine recommendations by chatting about them
The recommendations Mimir generates are a strong starting point, but your product intuition matters. Open a recommendation in the workspace panel and chat about it to adjust scope, shift priorities, or add constraints. The rationale updates live as you talk through changes.
Use impact and effort tags for prioritization
Each recommendation includes impact and effort estimates. Use the sort and filter controls to surface quick wins (high impact, low effort) — usually the best place to start.