1. CLAIM EXTRACT C1 - "Dara has had a genuinely excellent year" - TRAIT C2 - "is one of the most reliable people on the platform team" - TRAIT C3 - "She led the settlement retry work from design through rollout" - BEHAVIOUR C4 - "it has been rock solid since" - IMPACT C5 - "I have not had to think about that part of the system once since April, which says everything." - IMPACT C6 - "She is also the person other engineers go to when something is on fire" - TRAIT C7 - "she never makes anyone feel small for asking" - BEHAVIOUR C8 - "Her mentoring of the two new grads has been a quiet highlight" - BEHAVIOUR C9 - "both of them are shipping independently now" - IMPACT C10 - "I credit that largely to her" - IMPACT C11 - "She writes the clearest design docs on the team." - TRAIT C12 - "she could put herself forward more in the wider architecture forums" - BEHAVIOUR C13 - "where I think people would listen" - TRAIT C14 - "Overall rating: 5" - RATING 2. EVIDENCE MATCH C1 - "Jul - p95 ledger 410ms, unchanged. Deprioritised after retry slipped." - NARROWER - the record carries one goal delivered and one unmet, not a verdict over the whole year. C2 - "Aug 19 - covered my oncall over leave, no drama." - NARROWER - the record holds one reliable act, not a ranking against the platform team. C3 - "Apr 11 - retry shipped behind flag. 100% Apr 29." - SUPPORTED - with "Feb 3 - retry design doc v1 up. Rewrote twice after Anwar's comments." the record carries design through rollout inside the period. C4 - "May 22 - rollback, dup writes on partial settle. Caught by Ops, not us. Fixed May 30." - CONTRADICTED - "rock solid since" against that rollback and "Jul 8 - second rollback, smaller, config." C5 - the May 22 rollback line above, with "Jul 8 - second rollback, smaller, config." - CONTRADICTED - two rollbacks after April, the first found by Ops rather than by the team. C6 - no line found - UNSUPPORTED - no entry logs another engineer coming to her during an incident. C7 - "Peer form (Anwar): great engineer, review comments can land hard, have seen juniors sit on a PR rather than ask her again." - CONTRADICTED - "never makes anyone feel small for asking" against juniors sitting on a PR rather than asking again. C8 - "Grad Kelsi ramping, doing own tickets since Aug. Second grad Mo went to Owen's squad in Jun." - NARROWER - the record carries one grad against the goal "mentor one grad"; the second changed squads in June. C9 - the Kelsi and Mo line above - NARROWER - it names one grad doing her own tickets and calls her ramping; neither "both" nor "independently" is in it. C10 - the Kelsi and Mo line above - NARROWER - the addition is the causal credit, extended to a grad who moved to another squad in June. C11 - "Feb 3 - retry design doc v1 up. Rewrote twice after Anwar's comments." - NARROWER - one doc, rewritten twice after review comments; no other engineer's docs appear, so the ranking is not in the record. C12 - no line found - UNSUPPORTED - no entry names an architecture forum attended, missed or declined. C13 - UNCHECKABLE - how colleagues would receive her is potential, which no record holds; carried to section 6. C14 - no line found - UNSUPPORTED - the record sets goals and logs events and assigns no level. 3. RATING CHECK C14 - rating as written 5 - "5 - Consistently exceptional; sets the standard for others in the discipline." - SHORT - the only SUPPORTED row is C3, and nothing in section 2 reaches "sets the standard for others in the discipline"; that row sits better under "3 - Fully meets the expectations of the role." 4. DISPUTE RISK Eight lines carry dispute risk. C13 - "where I think people would listen" - "You are marking me on a reaction that has not happened, in a room the record never says I was in; there is nothing here I can disprove." C12 - "she could put herself forward more in the wider architecture forums" - "I asked for more scope and about the staff track in the June 1:1; no entry says I stayed quiet anywhere." C7 - "never makes anyone feel small for asking" - "Anwar's form says juniors sit on a PR rather than ask me again, and this review is the first place I see it - praise written over feedback nobody gave me." C4 - "rock solid since" - "There is a duplicate-write rollback on 22 May that Ops caught and a second on 8 July; if the review is wrong on that, the rest of it was not checked either." C5 - "I have not had to think about that part of the system once since April" - "Two rollbacks after April are in the same record, so this reads as written without opening it." C9 - "both of them are shipping independently now" - "Mo went to Owen's squad in June and Kelsi is logged as ramping; the sentence gives me an outcome the record does not show." C11 - "the clearest design docs on the team" - "No other engineer's docs are in this record, so this cannot be checked either way in calibration." C2 - "one of the most reliable people on the platform team" - "A ranking across the team rests here on one entry about covering an oncall." 5. REWRITE LIST C4 - narrow it to the record - "The retry redesign reached 100% on 29 April; a duplicate-write bug on partial settle was rolled back on 22 May, caught by Ops and fixed on 30 May, and a smaller config rollback followed on 8 July." C5 - cut it; the same two rollbacks contradict it. C7 - cut it; the peer form states the opposite and no line supports it. C12 - cut it, or narrow it to what is logged - "In our June 1:1 you asked about more scope and about the staff track." C1 - narrow it to the goals - "You delivered the settlement retry redesign; the p95 ledger goal stood at 410ms in July, unchanged." C2 - replace it with the behaviour logged - "You covered my oncall over my leave on 19 August with no drama." C11 - narrow it to the doc - "Your retry design doc went up on 3 February and you rewrote it twice after Anwar's comments." C8 - narrow it to one grad - "You mentored Kelsi, who has been doing her own tickets since August." C9 - fold into the C8 replacement and drop the second grad. C10 - fold into the C8 replacement and drop "largely". Below this line the repair is evidence rather than wording. C13 - nothing can carry a claim about how people would react; cut it or replace it after talking to whoever runs those forums. C6 - needs the engineers who escalated to her; no incident line names one. C14 - the 5 rests on a single SUPPORTED row; it needs evidence of setting the standard for others, gathered from the people who saw the work, or a different level. 6. NOT CHECKED Self-assessment - not in the record; the employee's own account of the cycle is absent. Peer feedback beyond Anwar's one-line summary - the form itself is not contained, and no other peer form appears. Retry design doc v1 and Anwar's comments on it - referred to on 3 February, not contained. June 1:1 notes - summarised in one line; what was agreed on scope or the staff track is not recorded. Previous cycle's review - not present, so "genuinely excellent year" has no baseline to be read against. Mentoring of Mo, January to June - no entry covers it either way. January, March, and anything after 19 August - the record says nothing about these stretches of the period. C13 - "where I think people would listen" - UNCHECKABLE; no record holds how colleagues would receive her.