Most promotion cases fail for the same reason: they describe the candidate rather than demonstrate them. "I'm ready for the next level" and "I consistently show strong technical judgment" are claims, and claims from an interested party carry very little weight with a committee that's never worked with you directly. What actually moves a promotion decision is evidence specific enough that someone with zero context could read it and reach the same conclusion independently.
Start collecting evidence before you need it
The biggest mistake is starting to assemble a promotion case the month before the cycle closes, trying to reconstruct six months of impact from memory. By then the details are gone — the exact metric improvement, the name of the person who said your design doc changed their approach, the specific decision you made under pressure that avoided an outage. Keep a running document, updated in the moment, not reconstructed later. When you ship something meaningful, resolve a hard incident, or get unsolicited praise from a peer, write it down immediately with enough detail that it's still legible in eight months.
Evidence beats adjectives
Every level's rubric uses words like "significant," "consistently," and "independently," and the temptation is to just restate those words back at the committee. That doesn't work because the committee's job is to judge whether the words apply — repeating them back is circular. Replace every adjective with a fact that implies it. Instead of "led a complex cross-team project," write "coordinated the migration across three teams, wrote the technical plan that became the reference doc for two other org migrations, and delivered it two weeks ahead of the original estimate." The second version doesn't need to claim leadership; the facts make the claim for you.
Quantify where you honestly can, but don't force numbers onto things that don't have them — a fabricated or stretched metric is worse than no metric, because a sharp committee member will notice the stretch and it undermines the rest of the packet. Qualitative evidence is fine when it's specific: a direct quote from a peer's feedback, a concrete description of a decision and its consequence, a before-and-after description of a process you changed.
Third-party corroboration matters more than self-report
A committee weighs a peer's unprompted praise far more heavily than your own description of the same event, because self-assessment has an obvious incentive problem and peer feedback doesn't. Actively collect this — ask people you've worked closely with for specific, written feedback while the work is still fresh in their memory, not vague endorsements after the fact. A manager or skip-level who can speak to your impact from their own vantage point, with specifics, is worth more in a packet than three more bullet points you wrote yourself.
Calibrate scope against the actual bar
Before writing anything, get an honest read on what the target level actually requires at your company — the written rubric is a starting point, but the real bar is often set by recent examples of people who were and weren't promoted. Ask your manager directly what's missing, not just what's present. A packet stuffed with evidence for the level you're already at, no matter how well-written, doesn't make a case for the next one; it just proves you're solid where you are. The strongest cases show a small number of examples that clearly demonstrate next-level scope, judgment, or impact, rather than a long list that mostly demonstrates competence at the current level with more items.
A worked failure mode
An engineer writes a promotion doc that lists every ticket and a vague "I improved reliability." The committee asks what broke, what changed, and who else could claim the same incidents. The packet has no before/after graphs, no named partners, and no unsuccessful bets. A peer who shipped a smaller, measured migration with quotes from partner teams gets the yes. The failure is a narrative without artifacts. Keep a running log: metrics, design docs, incident roles, mentee outcomes, and decisions you owned. Tie each claim to a link. Ask your manager quarterly whether the evidence matches the next level's rubric, not the week before the packet is due.
When this is the wrong tool
A promotion packet is the wrong tool to fix a broken leveling system or a manager who will not sponsor you; that is a different conversation (and sometimes a different job). Do not spend a quarter polishing slides instead of doing the work the rubric asks for. Evidence files are the wrong place for grievances. If the org promotes on tenure and relationships only, a perfect packet still may not move the needle—decide with eyes open. Use evidence when the process claims to be evidence-based and you can map work to the published bar.