promotion_case_building v1.0.0

---

name: promotion_case_building description: Build a promotion case that survives calibration: keep a year round impact log, turn project work into quantified business outcomes, map evidence to the next level guide, line up peer and stakeholder evidence, write the case document, prep the manager for the calibration room, and turn a not this cycle answer into a dated gap plan. Use when someone is aiming for promotion next cycle, writing a self review or promotion packet, helping a report get promoted, or recovering after a promotion was declined.

---

Promotion Case Building

Purpose

Produce a promotion packet that a stranger in a calibration room can defend without the candidate or the manager in the room. The packet rests on three things: a log kept all year, evidence mapped line by line to the next level, and corroboration from people outside the reporting line.

A promotion case is not a reward claim for hard work this year. Most calibration processes treat promotion as recognition that the person already works at the next level, sustained over time, in a role the organization actually needs. Build for that test. If the evidence does not meet it yet, this skill produces a dated gap plan instead of a weak packet. A weak packet submitted anyway costs the manager credibility in the room and makes the next attempt harder.

Audiences:

1. The employee building their own case and handing material to their manager.
2. The manager writing and defending the case for a report.
3. A peer or mentor reviewing a draft packet before it goes in.

Workflow

1. Pull the actual bar before collecting anything

Get these documents first. Without them the rest is guesswork.

• The level guide or competency framework for the current level and the target level, side by side.
• The promotion cycle rules for this org: who nominates, who decides, when calibration meets, whether a business need or open role at the next level is required, and whether the last rating gates eligibility.
• The candidate's last two reviews, including every development area named in them.
• Any prior promotion feedback, in writing if it exists.

Write the target level expectations as a short list of testable statements, for example "owns a multi team project end to end" rather than "shows leadership". Every later section maps to this list.

Stop and fix the foundation when: there is no level guide and no one can state what the next level does differently, the manager will not sponsor the case, or the last review names a gap the candidate has not addressed. In each case the output is a gap plan (step 7) with the missing piece as its first item, not a packet.

2. Run the impact log all year

The log is the system. Reviews written from memory lose to reviews written from a log, because memory keeps the last few weeks and drops the rest. US OPM guidance for supervisors makes the same point from the other side: keep records of accomplishments, feedback, and coaching throughout the period, document with specific examples, and use calibration to normalize across teams.

Cadence and rules are in references/impact-log-template.md. The short version:

• Add entries weekly, in the week the work happened. Ten minutes on Friday beats a weekend of reconstruction in review season.
• One entry per outcome, not per task. Include date, project, your specific action, the result with a number and a baseline, who else was involved and at what level, and a link to the artifact.
• Record glue work in the same log: mentoring, unblocking another team, design and documentation, interview loops, incident help, onboarding. Julia Evans' brag document structure covers exactly these categories, and they are the work most likely to vanish from a project list.
• Save praise in the moment. When someone thanks you in writing, copy the quote, the name, the role, and the date into the log. Asking for that quote nine months later rarely works.
• Never inflate. Evans' rule for brag documents applies here too: make the work sound exactly as good as it is. Calibration rooms test numbers, and one inflated figure poisons the true ones beside it.

3. Convert project work into business outcomes

For each candidate entry, rewrite it from activity to outcome using this shape:

• Situation: what was wrong, at what scale, and what it cost.
• Task: what you personally owned.
• Action: the decisions and steps you took, in "I" language, with the team's work named separately.
• Result: the change, as baseline to result, over a stated period, with the source of the number.

Worked conversions:

| Activity version (weak) | Outcome version (usable) | |---|---| | Led the checkout migration | Led the checkout migration off the legacy gateway between March and May. Failed payments fell from 4.1 percent to 1.8 percent of orders (payment dashboard, weekly average), and two partner teams unblocked their launches in June | | Improved onboarding docs | Rewrote the onboarding guide and added a first week checklist. New hires reached their first merged change in 3 days instead of 7 across the last four cohorts, per the team's onboarding tracker | | Helped with the incident | Ran comms for the 14 August outage: status updates every 30 minutes for 4 hours, postmortem published in 2 days. Support tickets about the outage dropped to zero after the second update |

Quantify honestly when the number does not exist. Options in order of strength:

1. Counts you can source: users affected, teams unblocked, hours returned, tickets closed, releases shipped.
2. Before and after from a system of record: dashboard, tracker, finance report. Name the source in the entry.
3. A range with the method stated: "roughly 6 to 8 hours a week across three teams, from their leads' estimates".
4. A named witness with no number: a quote from the person who received the benefit.

Never invent a baseline. If there was no baseline measurement, say the result without a delta and note that no baseline was captured. Reviewers trust that admission and distrust a suspiciously round improvement.

4. Map evidence to the next level

Build the alignment matrix before writing prose. One row per target level expectation:

| Target expectation | Strongest evidence (date, outcome) | Second source | Verdict | |---|---|---|---| | Owns multi team work end to end | Checkout migration, Mar to May, 3 teams, result above | Partner PM quote, 2 Jun | Met | | Coaches others to deliver | Mentored two engineers, both shipped first solo projects in Q2 | Mentee feedback in June review cycle | Met | | Sets technical direction | Wrote the payments roadmap doc, adopted Oct | No second source yet | Partial |

Rules:

• Two independent pieces of evidence per expectation is the working standard. One project carrying every row is a scope problem, not a writing problem.
• Mark a row Partial or Missing when it is. A packet that admits one partial row and names the plan for it reads as credible. A packet that claims everything is met invites the room to find the weak row themselves.
• Address every development area from the last review by name, with the evidence that it closed or an honest status. Calibration members remember old development areas longer than they remember accomplishments, and an unaddressed one becomes the discussion.
• Check sustained performance, not a spike. Evidence spread across at least two quarters, or the equivalent in your cycle, is far stronger than one strong month before nominations.

Run bin/packet_check.py on the draft packet to catch unquantified claims, missing second sources, and unnamed development areas before a human reads it.

5. Line up peer and stakeholder evidence

The manager's opinion is expected, so it carries the least weight in the room. Corroboration from outside the reporting line is what moves a borderline case.

• Pick three to five people: a peer on another team, a stakeholder who received the work, a skip level or partner lead who saw it firsthand, a mentee for coaching claims.
• Ask early, at least three to four weeks before the packet is due, and ask in writing so the reply is quotable.
• Ask a specific question, not "can you write me feedback". Use: "For my promotion packet I need evidence on the payments migration. Would you be willing to write two or three sentences on what you saw me own and what changed for your team?" Specific asks get specific answers.
• Quote people accurately and with permission, with name and role attached where your process allows it. Anonymous praise is discounted.
• Collect evidence about how the work was done, not only what shipped: collaboration, judgment under pressure, how disagreements were handled. Contested cases in calibration are usually contested on the how, so a packet with no how evidence has no defense ready.

Keep the raw replies with the log. Do not edit a quote beyond trimming length, and mark trims with ellipses.

6. Write the case document

Full outline in references/case-document-outline.md. A complete worked packet, with fictional people and numbers clearly marked as invented for illustration, is in references/worked-packet.md.

Shape:

1. Ask, one line: name, current level, target level, team, and the cycle.
2. Recommendation, two to three sentences: already operating at the target level since when, doing what scope.
3. Alignment matrix: the table from step 4, trimmed to the strongest evidence.
4. Three impact stories: the biggest outcomes in Situation, Task, Action, Result form, each half a page at most.
5. Corroboration: three to five short attributed quotes, each tied to a matrix row.
6. Trajectory: where the person was a year ago, what scope they took on since, in four to six lines.
7. Development areas: each prior area, its status, and its evidence.
8. Business need: why the role at the next level exists on this team now, if your process requires it.
9. Risks and honest gaps: one or two lines naming the weakest row and the plan for it.

Write it to stand alone. Assume the manager is not in the room, is ill, or gets two minutes instead of ten. Every claim should be understandable and checkable by a reader who has never met the candidate.

Length discipline: the core case fits in three to five pages plus attachments. Longer packets get skimmed, and skimmed packets lose their best evidence. Put artifacts behind links with a one line note on what each link proves.

7. Prep for calibration, then handle either answer

Prep with the manager before the meeting:

• Walk the matrix row by row and agree which row the room will attack. Prepare the extra example for that row in advance.
• Agree the one sentence version: "She has run multi team payments work at the next level since March, with results two partner teams will vouch for." If the case cannot be said in one sentence, it is not ready.
• List who else in the room has seen the work firsthand, so the manager knows where support will come from.
• Rehearse the how questions: a disagreement handled badly, a project that slipped, a piece of critical feedback. Prepare a factual answer for each, with what changed after.

If the answer is yes: get the effective date, the new expectations in writing, and what the first review at the new level will be judged on. Promotion moves the bar, and the first cycle at the new level starts at the decision, not at the next review.

If the answer is not this cycle, treat the debrief as the real deliverable:

1. Get the specific reasons in writing within days, while the room's reasoning is still available to the manager. "Not ready yet" is not a reason. Push for which matrix rows failed, and whether the blocker was evidence, scope, how, or business need. Each blocker has a different fix.
2. Sort each reason: real gap, evidence gap (did it, cannot prove it), or structural (no role, budget, headcount). Only the first needs new work. The second needs better logging and witnesses. The third needs a team or role conversation, not more effort.
3. Turn the reasons into a gap plan: one row per gap, the concrete work or evidence that closes it, the owner (candidate, manager, or both), a check date, and the target cycle. Template in references/case-document-outline.md.
4. Agree what the manager will do differently: sponsor earlier, put the candidate on the scope that is missing, or introduce them to the people who need to see the work.
5. Set a mid cycle check, roughly halfway to the next cycle, to test whether the evidence is actually accumulating. Waiting until the next nomination to discover the gap is unchanged wastes a full cycle.
6. Keep the tone factual in all of this. A declined case with a written gap plan is a stronger starting position next cycle than a fresh case built from memory.

Output Contract

Deliver, in this order:

1. Bar summary: the target level expectations as testable statements, plus the cycle rules that gate the case.
2. Impact log: the running log in the template format, or a cleaned extract covering the review period.
3. Alignment matrix: every target expectation with its evidence, second source, and an honest Met, Partial, or Missing verdict.
4. Case document: the nine part packet from step 6, or a gap plan when the matrix shows the case is not ready.
5. Evidence pack: the attributed peer and stakeholder quotes, with the raw replies preserved.
6. Calibration prep note: the one sentence case, the row most likely to be attacked, and prepared answers for the how questions.
7. After the decision: either the new level expectations in writing, or the dated gap plan with owners and a mid cycle check.

State plainly which of these you produced and which are missing and why. Never present a gap plan as if it were a submitted packet, or a draft packet as if it had been filed.

Operating Rules

• Verify every number in the packet against its source before it goes in. A figure nobody can trace gets cut, not rounded into plausibility.
• Never invent numbers, dates, quotes, ratings, or feedback. Unverifiable material is cut or labeled as an estimate with its method stated.
• Never edit a peer quote to change its meaning, and never write a quote for someone and ask them to sign it. Ask for their words, accept what they give.
• Keep the candidate's own contribution separate from the team's result. Claiming team outcomes as solo work is the fastest way to lose a room that includes those teammates' managers.
• Do not hide a prior development area, a slipped project, or a piece of critical feedback that the room is likely to know about. Surfaced by you, it is context. Surfaced by the room, it is a credibility problem.
• Respect confidentiality. Peer feedback, ratings, and calibration discussion contents follow your org's rules. Share a person's own feedback with them as policy allows, and never quote what was said about another candidate in calibration.
• Check for bias in your own packet before filing. Rate each matrix row on evidence alone, watch for recency (all evidence from the last quarter), halo (one strong project lifting unrelated rows), and similar to me reasoning. OPM guidance names recency, halo, and favoritism as the biases calibration exists to correct, so write the packet as if the room will test for them, because it will.
• No promises. Neither the employee version nor the manager version of this work guarantees an outcome. Say what the case does: it makes the evidence hard to argue with. The decision belongs to the process.
• Follow the org's actual process when it conflicts with this workflow. Level guides, eligibility rules, and deadlines differ by employer and change over time. Pull the current versions instead of assuming any company's rules.