How much repeat-feedback work are you modeling?
Three minutes of inputs. The math is yours, not a sales pitch.
Your team
Numbers update live. Tweak anything that doesn't match your team.
Share of comments that re-explain things the team already knows (style, naming, prior decisions, repeated bugs).
Combined: reviewer reads + writes, author reads + reacts.
Your annual review-effort waste
Estimate based on the inputs at left.
From your repeat-feedback rate.
Conservative estimate: 60% of repeat comments. We chose 60% because team review memory can surface the "we already decided this" class before repeat work — not bugs that need a human eye.
Compared against $240 / seat / year (Team list anchor; final pricing on /pricing).
Annual Team plan assumption used for the multiple.
Why these numbers hold up
- 01 Real review memory, not a generic policy. difflore extracts rules from your team's actual PR comments, so the model targets feedback you've already given — not a pre-baked checklist.
- 02 OSS local runtime, BYOK key works. The CLI is open source and runs against your own model key. The cloud plan only kicks in when you want shared memory and review automation — no vendor lock-in on the modeled value above.
- 03 Conservative on the repeat-signal model. The 60% model covers repeat feedback only, where surfacing the prior decision is the whole job. Bug-finding upside is on top.
Model the repeat-feedback tax.
Connect a repository, import your past reviews, and difflore starts surfacing what your team already decided.
These numbers are estimates. Actual savings depend on team size, repeat-feedback rate, and corpus volume. The Team seat-cost anchor is a list-price reference; current pricing lives on /pricing.