Promotion Packets in Engineering: How to Write, Review, and Calibrate Without Losing Trust

60-Second Summary

A promotion packet is a written case. Its job is to make the decision easy and defensible, not to flatter the candidate.
Every packet must answer four questions: scope, impact, behaviours, and counter-evidence.
The manager's review is a second draft, not a thumbs-up. Managers who 'submit and hope' lose more promotions than they win.
Calibration meetings are where unfairness is born or eliminated. Run them with a written rubric, a designated note-taker, and an explicit bias check.
Publish redacted past packets internally. Secrecy is the single largest source of perceived unfairness.

If you survey engineers across 50 companies and ask which HR process generates the most cynicism, the answer is almost always the same: promotions. Not compensation, not performance reviews — promotions. The reason is structural. The promotion decision aggregates a year of ambiguous evidence, in a room the candidate is not in, judged against a rubric that often only exists implicitly. A well-written promotion packet, reviewed thoughtfully and calibrated transparently, is the single highest-leverage process change an engineering HRBP can make.

Why packets matter

A promotion packet does three things at once: it organises evidence, it forces the candidate to articulate their own case, and it produces an artefact the calibration room can argue about. Without a packet, the discussion defaults to whichever manager has the strongest narrative and the loudest voice. Google, Meta, Stripe, Airbnb, Microsoft, and most large engineering orgs run on packets for exactly this reason.

Plain-English definition

A promotion packet (also: 'promo doc', 'self-review', 'case for promotion') is a 5–15 page written document making the case that an engineer should be promoted to the next level. It is written primarily by the candidate, edited by the manager, and read by the calibration committee.

What a packet contains

The four questions every packet must answer

1
Scope
How big is the work? Number of engineers / teams / users / dollars affected. What's the time horizon? Is it growing or shrinking?
2
Impact
What measurably changed because of this person's work? Revenue, latency, headcount, hire rate, defect rate. Distinguish 'I contributed to' from 'I personally caused'.
3
Behaviours at level
Evidence of the leadership behaviours the next level requires — written communication, technical mentorship, cross-team influence, decision-making under ambiguity. Pulled directly from the rubric.
4
Counter-evidence
Where is the candidate still operating at the current level, not the next? An honest packet names two or three. This is the section that earns the room's trust.

Standard packet structure

Section	Length	Purpose
Summary	1 paragraph	The 30-second pitch. Level, scope, top 2 outcomes.
Scope and role	0.5 page	What the candidate owns, who they work with.
Top 3–5 projects	2–4 pages	One per project: situation, action, result, evidence. Hyperlinks to artefacts.
Leadership and behaviours	1–2 pages	Rubric line items with specific examples.
Growth areas / counter-evidence	0.5 page	What the candidate is working on next.
Peer and partner quotes	0.5 page	3–5 short quotes from named partners (PM, designer, partner team eng).
Appendix	as needed	Links to design docs, RFCs, PRs, post-mortems, public talks.

Writing the packet (engineer view)

Start 6–8 weeks before submission. Block 1 hour per week.
Open with the rubric. Read the next level's expectations and highlight every line where you have evidence.
List every meaningful project from the review period. Rank by impact, not effort.
For each top project, write a STAR paragraph: Situation, Task, Action, Result. Include numbers wherever defensible.
Use 'I' for what you personally did. Use 'we' for team outcomes. Never inflate.
Collect 3–5 short quotes from named partners. A two-line quote from a director carries more weight than a paragraph from a peer.
Write the counter-evidence section honestly. The room will find these things anyway — naming them yourself converts a liability into credibility.
Read the packet out loud. Cut anything you can't defend in person.

The 'modest engineer' tax

Engineers who under-describe their work get promoted less often than equally strong engineers who write a clear packet. This is a real, documented effect (see Google's Project Aristotle / re:Work analyses, and the body of research on self-promotion gendered differences — Babcock & Laschever). Managers must coach for clarity, not modesty.

Manager review and second draft

Submitting a packet without a thorough manager edit is a 30–40% drop in promotion odds in most calibration systems we have data on. The manager's job is not approval — it is editorial.

The manager's three passes

1
Pass 1 — Evidence
Every claim has a hyperlink or named witness. Anything that doesn't is cut or sourced.
2
Pass 2 — Level fit
Re-read against the rubric. Mark sections that show next-level vs current-level evidence. If the next-level evidence is thin, say so before submission — don't ambush the candidate in calibration.
3
Pass 3 — Defensibility
Imagine the most skeptical director in the room. Can you defend each project's scope claim? If not, soften the language now, not after.

Running the calibration meeting

Calibration is where unfairness is born or eliminated. The single largest predictor of fairness is structure.

A calibration meeting that holds up

1
Pre-read
Every committee member reads every packet before the meeting. No reading during the session.
2
Designated note-taker (not a manager)
Captures the rationale for each decision. The note becomes the feedback artefact.
3
Rubric on the wall
Literally on screen. Every decision references a specific rubric line.
4
Sponsor presents, others challenge
Manager presents in 5 minutes. 10 minutes of structured challenge. Then a vote.
5
Explicit bias check
Before any vote, the chair asks: 'What evidence are we relying on here that we wouldn't accept from a different candidate?' Awkward, necessary.
6
Outcome and rationale captured in writing
Three sentences minimum. Becomes the feedback for the candidate.

Bias checks that actually work

Aggregate promotion rates by gender, ethnicity, tenure, and team. Look at the multi-year trend, not just this cycle.
Spot-check 5 random declined packets vs 5 approved packets in the same cycle. Was the rubric applied consistently?
Watch for 'too soon' language disproportionately applied to candidates from under-represented groups — it is the single most cited finding in Lean In's and McKinsey's joint Women in the Workplace research.
Watch for vague language ('strong communicator', 'good cultural fit') that has no rubric line. Strike it from the discussion.
Track the gap between manager nomination rate and committee approval rate by demographic. Persistent gaps are the data.

Decisions and feedback

Every candidate — approved or declined — receives a written summary of the rationale within one week. For declines, the note answers two questions: what specifically would the next packet need to show, and what timeline is realistic? Vague decline feedback ('keep growing!') is the second largest source of cynicism after secrecy.

Anti-patterns

No written rubric — calibration becomes a personality contest.
Manager submits the packet without the engineer writing it — destroys the engineer's ability to reflect and own their narrative.
Calibration meetings closed to scrutiny — breeds rumour culture.
'Quotas' or unspoken caps on number of promotions per cycle — silently distorts every decision.
Surprise declines without prior coaching — destroys trust in the manager relationship.
Same packet template for ICs and managers — measures the wrong things for one of them.

Templates and examples

Publish 3–5 redacted past packets internally each cycle. Yes, it feels uncomfortable. The trust gain dwarfs the cost. Lara Hogan's resources and StaffEng.com host several public examples; use them as anchors, not as the company-specific rubric.

Monday-morning checklist

Confirm a written rubric exists for every level you promote to. If not, this is the first project.
Audit last cycle's declines. How many came with written feedback within one week?
Publish 3 redacted past successful packets internally.
Train calibration chairs on the bias-check question.
Aggregate promo rates by demographic and review with the exec team annually.

FAQ

Frequently asked questions

Should peers see the packet?

Some companies (Stripe early on) allowed peer feedback inline; most have moved to manager-mediated peer quotes for fairness. Decide explicitly and document it.

What about engineers on parental leave?

Prorate the review window. Never penalise leave time. Several large companies have policies in writing — make yours explicit.

How long should the cycle take?

From self-write start to decision, 6–8 weeks. Anything shorter rushes evidence gathering; anything longer creates anxiety overhead.

Can we skip packets for obvious promotions?

Don't. The packet's value is in the process — both for the engineer and for the institutional record. 'Obvious' promotions also need defensible artefacts.

References

Lara Hogan — Promotion packets and resources — Lara Hogan
StaffEng — promotion packet examples — StaffEng
Lean In & McKinsey — Women in the Workplace (annual) — Lean In
Babcock & Laschever — Women Don't Ask (2003) — Princeton University Press

Written by Pawan Joshi.Sources cited inline.

First published 15 Jun 2026See site changelog →