Impact Evidence, Calibration, and Promotion
Senior-to-Principal Growth OS
Overview
Evidence makes impact inspectable
Impact evidence connects a meaningful changed outcome to the engineer's contribution, the surrounding context, and corroboration. It helps reviewers inspect a claim without pretending one person caused a complex organizational result alone.
Frameworks from Monzo, Dropbox, and CircleCI emphasize impact and widening scope in different language. Use your organization's framework as the decision context; use an evidence system to make the underlying work clear.
Keep an impact ledger
An impact ledger is a lightweight running record of outcomes and contribution. Update it while evidence is fresh rather than reconstructing a year from calendar invitations and memory.
| Field | What to record | Example |
|---|---|---|
| Problem and baseline | Why the work mattered before intervention | Recovery took 95 minutes; ownership unclear across four services |
| Scope and ambiguity | Systems, teams, duration, unknowns | Four teams; no shared reliability model |
| Contribution | Decisions, direction, alignment, enablement | Framed options, created owner model, coached two leads |
| Collaborators | Who did what | Service leads implemented; the site reliability engineering (SRE) team supplied incident data |
| Outcome | Quantitative and qualitative change | Recovery median fell to 28 minutes; owners respond independently |
| Durability | What persists beyond the project | Runbook, ownership, measures, and trained leads |
| Corroboration | Links, data, and relevant observers | Dashboard, RFC, postmortem, three stakeholder observations |
| Context | Constraints, changes, and uncertainty | One service excluded after compliance review |
The ledger is not a diary of everything completed. Record only work that teaches something about impact, capability, or the quality of an opportunity.
Write claims without erasing collaborators
Strong attribution is precise about the engineer's distinctive contribution and equally precise about shared execution. “I did everything” is usually implausible at organizational scope; “the team did it” can hide real technical leadership.
Use this four-part claim:
Outcome: [measured or observed change]. My contribution: [specific decisions, mechanisms, or leadership]. Shared work: [named collaborators and their contribution]. Evidence: [data, artifact, observer, and durability].
Example:
Outcome: Four services cut median recovery from 95 to 28 minutes over two quarters. My contribution: I framed the ownership problem, aligned directors on one operating model, and coached two service leads through adoption. Shared work: the leads implemented service changes and SRE built the measures. Evidence: incident dashboard, approved operating model, postmortems, and independent response by the new owners.
This claim is more credible than either “led reliability” or “helped the team.”
Collect several kinds of evidence
No single metric captures Staff+ work. Combine outcome, behavior, artifact, adoption, and relevant-observer evidence, then state uncertainty and counterevidence.
| Evidence type | Useful example | Limitation |
|---|---|---|
| Outcome | Reliability, cost, time, customer, or delivery change | Attribution may be shared or delayed |
| Behavior | Observed framing, dissent handling, delegation | One observation may not show a pattern |
| Artifact | Decision, strategy, interface, operating model | Production does not prove adoption |
| Adoption | Teams use and sustain the direction | Compliance can be mistaken for value |
| Talent leverage | Other engineers lead independently later | Causality is difficult to isolate |
| Counterevidence | A team rejected the path for valid reasons | Requires psychological safety to record |
Evidence quality rises when sources are close to the work and relevant to the claim. Executive praise may show communication or strategic confidence; it does not prove technical adoption inside teams.
Build a concise calibration packet
Calibration is a process in which leaders compare evidence against shared expectations to improve consistency. A concise packet helps reviewers reason; a long advocacy document can bury weak claims under volume.
Recommended packet:
- Proposed role and the organization's need for it.
- Two or three strongest impact claims across time.
- Direct mapping to the most relevant local expectations.
- Scope, ambiguity, leverage, and durability.
- Collaborators and corroborating artifacts or people.
- Material gaps, changed context, and counterevidence.
- Manager recommendation and remaining uncertainty.
Do not map every sentence to every ladder cell. Dropbox explicitly says its framework is not a checklist, and that warning generalizes: reviewers need the pattern and consequences of the work, not a scavenger hunt.
Separate development review from promotion decision
A development review asks what capability is becoming reliable and what opportunity should come next. A promotion decision asks whether sustained evidence matches a needed role under the organization's process; related inputs do not make these identical conversations.
| Development review | Promotion decision |
|---|---|
| Can happen every quarter | Happens under company policy and timing |
| Optimizes learning and opportunity | Evaluates role need and sustained evidence |
| Can proceed despite incomplete evidence | Requires sufficient evidence for a consequential decision |
| Ends with next experiment and support | Ends with decision, rationale, and next steps |
GitLab's career guidance states that progression beyond Senior requires both demonstrated ability and organizational need rather than time in role alone. Completing this course, a contract, or one high-profile project does not guarantee promotion.
Use fair, candid language
Promotion conversations fail when managers promise what they do not control, hide the real standard, or keep moving expectations after results arrive. Engineers also need room to challenge evidence without treating every unfavorable observation as bad faith.
When evidence is not sufficient:
“I support your growth toward Principal, but I do not yet have sustained evidence of organization-wide problem selection. The strongest evidence is multi-team domain direction. I will not invent a deadline or promise a decision. I propose this real strategy assignment, the support listed here, and a review of evidence and role need on this date.”
When organizational need is missing:
“Your Staff-scope evidence is strong. The organization does not currently have the needed role or scope for this promotion. That is an organizational constraint, not a hidden performance gap. I will document the distinction, discuss realistic options, and not ask you to perform an indefinite unpaid version of a nonexistent role.”
Engineer response:
“Please separate demonstrated strengths, missing evidence, and organizational constraints. Which observations support each conclusion, and which actions do you own?”
Audit calibration for bias
Fair calibration checks whether comparable evidence receives comparable interpretation and whether opportunity access shaped the evidence pool. It does not assume perfect objectivity is possible.
| Audit question | Risk detected |
|---|---|
| Are adjectives anchored in behavior and effect? | Personality and style bias |
| Do we value prevention and enablement beside rescues? | Visibility bias |
| Did the person receive authority and consequential opportunity? | Opportunity bias |
| Are collaborators credited consistently? | Solo-hero bias |
| Do communication norms favor one culture or temperament? | Style conformity bias |
| Is counterevidence applied equally to familiar and unfamiliar people? | In-group bias |
| Are standards stable across the review period? | Moving goalposts |
Managers should compare the opportunity log with promotion evidence. A thin packet may reflect limited access, not limited potential; that does not manufacture evidence, but it changes the development response.
Key takeaways
Promotion evidence should make a sustained impact pattern inspectable while preserving context and uncertainty. Honest evidence is useful even when it does not produce the hoped-for decision.
- Keep a lightweight impact ledger throughout the work.
- Connect baseline, contribution, shared work, outcome, durability, and corroboration.
- Combine evidence types and include counterevidence.
- Build a concise packet around the strongest sustained claims, not every ladder cell.
- Separate capability development, demonstrated evidence, organizational need, and promotion decision.
- Never guarantee promotion from a course, assignment, tenure, or completed checklist.
Practice: draft one impact claim and red-team it
This exercise produces one calibration-ready claim. Ask a collaborator and someone outside the project to review it for different weaknesses.
- Draft the four-part outcome, contribution, shared-work, and evidence statement.
- Add baseline, scope, ambiguity, and durability.
- Name one material uncertainty or counterexample.
- Ask a collaborator: “Is credit accurate and complete?”
- Ask an outsider: “Can you understand why this mattered and what I contributed?”
- Remove activity detail that does not support the claim.
Checklist:
- [ ] Outcome is meaningful, not merely completed activity.
- [ ] Contribution uses observable verbs and does not claim solo causality.
- [ ] Collaborators receive specific credit.
- [ ] Evidence is linked or named and spans enough time.
- [ ] Organizational context and uncertainty are explicit.
- [ ] Claim maps to local expectations without treating them as a checklist.
Sources
These practitioner frameworks ground impact and career-process claims; the ledger, packet, scripts, and audit are course syntheses.