The engineering manager promotion criteria decision record: why the criteria model you chose determines your manager pipeline quality and your IC-to-manager transition failure rate

A 28-person developer tools company building a CI analytics platform reached a point in its growth where two of its four engineering teams were running without a dedicated engineering manager. The engineering director was covering both teams directly and was spending 60% of his time on management work that should have been delegated — 1:1s, sprint ceremonies, performance conversations, recruiting screens. The company needed to promote an internal candidate. The promotion decision, discussed in a two-hour leadership meeting, used a single implicit criterion: who is the most respected engineer on the team? The answer was clear and unanimous — a senior engineer who had been with the company from the start, had architected three of the four core pipeline modules, was the first reviewer most engineers tagged in code review, and whose technical opinion had shaped every significant architecture decision in the prior two years. The promotion was offered, accepted, and announced in the same week. No criteria framework. No management readiness assessment. The decision was: best engineer becomes manager.

Six months into the role, the new manager's direct reports were surfacing concerns in skip-level conversations with the engineering director. The manager was running 1:1s but treating them as engineering design discussions rather than as development conversations — the agenda was whatever the direct report was currently building, not the direct report's growth, concerns, or blockers. Two direct reports reported feeling micromanaged: the manager was reviewing and often rewriting their pull requests with the same intensity the manager had applied as a high-output IC. A sprint crunch in month four had produced two consecutive weeks where the manager cancelled 1:1s to write code — the manager's instinct under pressure was to contribute individual output rather than to protect the team's capacity and morale. When a direct report showed early signs of disengagement — shorter Slack responses, declining to take ownership of new tasks, arriving late to standups — the manager did not initiate a performance conversation; the situation was escalated to the engineering director three months after the first signals appeared, by which point the direct report had already accepted an offer at another company. The manager self-described the challenge in a skip-level conversation at month seven: "I keep feeling like the right thing to do is to fix it myself, and I know that's wrong, but I don't know what the right move is instead." The manager resigned at month nine, citing "this role isn't the right fit for me right now," and returned to IC work at a different company. The engineering team had lost its strongest technical contributor and still needed a manager for both teams — the hiring process that followed took four months and produced an external hire at a significantly higher comp band than the internal candidate had been earning.

The promotion decision had selected for the most visible IC excellence signals — architectural contribution, code review quality, institutional knowledge, peer respect — none of which predicted the management skills the role required: comfort with ambiguity in people problems rather than in technical problems, the ability to hold someone accountable for an outcome without controlling how the outcome was achieved, and the preference for developing other engineers' capabilities rather than for exercising one's own. Connect this to the engineering career ladder decision record: the career ladder model determines what promotion evidence is collected and how it is evaluated, and a ladder that does not have an explicit branch at the IC-to-manager junction — with distinct criteria for the management track rather than a continuation of the IC rubric — will produce promotion decisions that are made against the IC criteria by default, because those are the criteria that exist; the manager promotion decision inherits whatever assessment process the company has, and if the assessment process was designed to evaluate IC seniority, it will select for IC seniority regardless of what the promotion committee believes it is selecting for.

A 45-person B2B SaaS building a contract lifecycle management platform had grown its engineering organization to six teams over three years, and had promoted one engineer per team to engineering manager as each team formed. The promotions were all internal and all based on seniority and team tenure — the engineer who had been on the team the longest, who other engineers came to first with questions, and who the previous manager had recommended before departing. The company had no formal promotion criteria for the management track, but the informal criteria were legible to anyone in the engineering org: tenure, technical credibility, and manager recommendation. Over eighteen months, a pattern emerged in the engineering leadership team's monthly calibration sessions. Four of the six engineering managers were consistently delivering acceptable team output but showing no coaching signal: their 1:1 notes, reviewed quarterly by the engineering director, were status updates about current sprint work rather than development conversations about direct reports' growth, concerns, or career goals. One manager had not updated their 1:1 notes in two quarters. Three managers had zero documented performance conversations despite the engineering director's direct observation that two of their teams had engineers operating below expectations. Two managers had mentioned in separate skip-level conversations with the VP of Engineering that they missed writing code and were thinking about returning to an IC role.

The VP of Engineering conducted a structured listening tour — twelve conversations with engineering managers and senior ICs across the six teams over three weeks. The finding was consistent and uncomfortable: most of the engineering managers had not pursued the manager role because they wanted to manage. They had pursued it because management was the only career path available above senior engineer at the company. The IC ladder terminated at senior engineer; the only way to grow title and compensation further was to move into management. Three managers described the promotion decision explicitly as a career optimization rather than a role preference: "I didn't want to manage, but there was nowhere else to go." Two managers described feeling trapped: they had been in the manager role for two years, their compensation had increased, and they did not see a path back to IC work at the same comp level. One manager described the role's emotional texture precisely: "I do the work, and it's fine, but I don't find any of it interesting. I was interested in the technical problems. I'm not interested in the people problems." The company's manager layer was largely populated by engineers for whom management was an instrumental choice rather than an intrinsic preference, and instrumental managers — managers who are doing the job for career optimization reasons rather than because they find the work of management genuinely engaging — produce adequate management but not good management, and adequate management in a fast-growing company is a ceiling on team performance that is invisible in the sprint velocity data until the company's hiring and retention start reflecting what it is like to be managed by someone who would rather be doing something else.

The structural fix the company implemented was a deep IC ladder: staff engineer and principal engineer tracks with titles and compensation bands set at genuine parity with engineering manager and senior engineering manager, respectively. The immediate effect was not a flood of managers returning to IC work — the managers who had been in role for two years had legitimate value as managers, and most of them stayed — but a visible change in who was pursuing the management track in subsequent promotion cycles. The engineers who began the management readiness process after the staff track was introduced were self-selected for management motivation rather than for career optimization, and the first cohort of managers promoted under the dual-track model produced a measurably higher rate of proactive coaching conversations in their 1:1 notes and a lower rate of skip-level concerns about management quality in the first 12 months post-promotion. Connect this to the engineering manager span of control decision record: the manager motivation model and the span of control model are structurally coupled — a manager who finds the work of management genuinely engaging and has the capacity and support to do it well can manage 6 to 8 direct reports effectively; a reluctant manager who is performing management as a career obligation rather than as a genuine professional investment will struggle with 4 direct reports because the management work that produces good outcomes — proactive development conversations, early performance interventions, recruiting investment — requires discretionary effort that a reluctant manager will not reliably contribute; the span of control model assumes a certain level of management engagement, and a promotion criteria model that produces reluctant managers makes the span assumptions in the span of control model invalid regardless of what the org chart says.

A 65-person enterprise SaaS building a financial compliance platform had, after two painful experiences with IC-to-manager promotion failures in its first three years, invested in designing explicit management readiness criteria. The company's VP of Engineering had documented the criteria in a four-page promotion guide: the management candidate must have demonstrated coaching ability through a structured shadow experience (at least 90 days of co-leading a direct report's performance development with their current manager), must have led a cross-functional initiative with a named deliverable and a timeline, must have a documented instance of delivering difficult feedback to a peer engineer with a positive outcome, and must have interviewed and screened at least four engineering candidates to demonstrate familiarity with the hiring bar. The criteria were defensible, grounded in management work rather than IC excellence, and had been reviewed by an external engineering leadership advisor. The first cohort of five engineers assessed against the new criteria produced three promotions; the criteria process had correctly filtered out two candidates who met the IC excellence bar but did not meet the management readiness bar.

Twelve months after the first cohort's promotions, three of the five promoted managers had experienced formal performance interventions. The interventions were not for the same reasons as earlier promotion failures — the new managers were not reverting to IC behavior or avoiding 1:1s. The problems were more specific: one manager had been unable to navigate a conflict between two senior direct reports over project ownership for four months before escalating to the VP of Engineering, by which point one of the direct reports had downleveled their engagement; a second manager had not produced a development plan for any of their five direct reports in six months despite being aware that two engineers were at risk of exit for growth-related reasons; a third manager was conducting effective 1:1s but had no system for tracking the commitments made in those conversations, producing a pattern where the same issues were raised session after session with no resolution because the manager had no mechanism for following through. The criteria had correctly identified management readiness — all three managers had the instincts and motivation the criteria were designed to select for — but the promotion decision had been treated as the conclusion of the selection process rather than the beginning of a structured transition. The managers had been handed teams on day one with no first-90-days support model, no assigned senior management mentor for the transition period, and no explicit guidance on which management practices to focus on sequentially rather than all at once.

The post-mortem produced a transition support model that the company had not had: a 6-month structured ramp for new managers with a weekly check-in cadence with the VP of Engineering for the first two months, a standard first-90-days curriculum that sequenced the management practices new managers should develop (diagnostic 1:1s before development 1:1s; development conversations before performance conversations; performance conversations before delegation), and an explicit handoff protocol that required the new manager to have zero IC deliverables by end of week two — complete IC handoff was a precondition for starting the ramp, not a goal to work toward. The key insight from the post-mortem was structural: the criteria correctly answered "is this engineer ready to be a manager?" but the promotion decision's implicit assumption — that the answer to that question was sufficient to produce an effective manager — was wrong. Being ready to be a manager and being an effective manager from day one are different things. The gap between them is bridged by structured transition support, and a company that designs excellent promotion criteria but no transition support model has solved half of the problem. Connect this to the engineering onboarding ramp decision record: the IC onboarding ramp problem and the new manager transition problem share the same structural failure mode — the hiring or promotion decision is treated as the endpoint of a selection process rather than the beginning of a structured capability transfer; in both cases, the cost of an absent ramp is paid not by the person who was hired or promoted but by the people who are now working with them and depend on their effectiveness; in both cases, the ramp owner is the senior person responsible for the new hire's or new manager's success, and a ramp without a named owner is a ramp that is maintained by the new person's resilience and self-sufficiency rather than by a system designed to accelerate their effectiveness.

Structural properties set by the engineering manager promotion criteria decision

Two structural properties are determined by the founding decisions about promotion criteria model, calibration process, decision structure, and transition support. Neither is visible in the quarterly OKR data or the team health pulse survey until manager pipeline quality constraints begin limiting the company's ability to scale engineering teams, or IC-to-manager transition failure rates begin accumulating in the form of manager attrition, team delivery degradation, and the chilling effect on future internal promotion candidates who observe the transition's visible struggles.

Property 1: Manager pipeline quality. The criteria model determines whether the manager pool reflects genuine management capability — the ability to develop other engineers, hold teams accountable for outcomes, navigate people complexity, and represent the engineering team's capacity and constraints accurately upward — or reflects accumulated IC seniority packaged into a management title. A criteria model that selects for IC excellence produces a manager layer that is strong at technical judgment and weak at the structurally different work of management, which is invisible in the hiring data but surfaces in skip-level conversations over 6 to 12 months as direct reports describe feeling micromanaged, feeling unsupported in their development, or feeling managed by someone who is primarily interested in the technical problems their team is solving rather than in the people solving them. Manager pipeline quality has a compounding effect: a manager who is struggling in the role is not developing the next generation of technical leads and senior engineers who will eventually be the management candidates for the next growth phase; a company that promotes on IC excellence and experiences high new manager transition failure rates is depleting its internal management pipeline twice — once by promoting the IC into a management role where they struggle, and once by leaving the IC work they were doing undone, which requires backfilling from outside the company at a higher cost and with a longer ramp than an internal successor would have required. The pipeline quality property becomes most visible when the company needs to scale from 50 to 150 engineers: a company with a strong internal promotion criteria model can promote most of its new managers internally, at a cost and ramp that reflects institutional knowledge; a company that has been promoting on IC excellence and experiencing 40% new manager attrition in the first 18 months is a company that will fill most of its management capacity through external hiring, at a higher cost, with a higher ramp, and with the organizational disruption that comes from bringing management philosophy from outside rather than developing it internally. Connect this to the engineering hiring bar decision record: the internal promotion criteria model and the external hiring bar are complements in the management pipeline — the internal criteria determine what the company grows internally, and the external hiring bar determines what the company imports; a company that promotes poorly from inside will increasingly rely on external management hiring, and the external management hiring bar will then need to be calibrated against a management team that was built partially through failed internal promotions, which is a harder calibration problem than calibrating an external bar against a management team built through deliberate criteria.

Property 2: IC-to-manager transition failure rate. The transition failure rate is jointly determined by how accurately the promotion criteria predict management readiness rather than IC excellence and by the quality of the transition support model applied after the promotion decision. The failure rate is not visible in the performance review data at the time of the failure — a new manager who is struggling in month four will appear, in most performance systems, as a new manager who is still ramping, because most companies do not apply formal performance frameworks to new managers in their first two quarters. The failure becomes visible later: in skip-level conversations where direct reports begin expressing concerns about management quality; in the manager's own skip-level conversations where the manager begins describing the role's difficulty in terms that suggest they are not finding the management work engaging rather than that they are developing through normal early-manager challenges; and in the team's delivery and retention metrics, where a struggling manager produces a specific degradation pattern — slower decision-making because the manager is not comfortable holding ambiguity, higher attrition among senior ICs who have other options and whose standard for management quality is higher, and a project ownership distribution where the manager has retained the highest-stakes technical decisions rather than delegating them to senior direct reports. The compounded cost of a transition failure has four components. First, the direct cost: the manager's lost IC productivity during the period of struggling before the intervention, which is typically 6 to 12 months at a senior engineer's fully-loaded compensation. Second, the team cost: the direct reports' experience of being managed by someone who was not equipped to manage them, which produces its own attrition signal and requires its own recovery work after the manager situation is resolved. Third, the pipeline cost: the management slot that must be refilled, either through an external hire or through a new internal promotion cycle, at full recruiting and onboarding cost. Fourth, the chilling effect: the engineers on the team who observed the transition's struggles become the primary data source for other ICs considering the management track, and a visible transition failure produces a 12 to 18-month suppression in internal management interest that the company must actively work to reverse. Connect this to the engineering manager span of control decision record: the transition failure rate and the span of control model interact directly — a new manager on a 4-person team has more margin for transition struggles than a new manager on a 7-person team, because the complexity of the people management work scales with team size and the most common transition failures involve the new manager being overwhelmed by the volume of management decisions before they have developed the judgment to handle them efficiently; a span of control model that assigns new managers to teams at the upper end of the healthy range before they have completed a full management ramp is a model that is increasing the transition failure rate by reducing the margin for the learning that the ramp requires.

The engineering manager promotion criteria ADR: five sections

Section 1: Promotion criteria model and readiness signals. The promotion criteria model specifies what constitutes a management readiness signal as distinguished from an IC excellence signal, and what evidence is required before the criteria can be assessed. The fundamental distinction is between signals that predict IC performance and signals that predict management performance — these are different and frequently anti-correlated in the absence of deliberate selection. IC excellence signals are well-documented in most engineering organizations: code review quality, architectural contribution, output velocity, technical mentorship of more junior engineers, and cross-team problem-solving. Management readiness signals require deliberate observation and cannot be inferred from the IC performance record: demonstrated coaching that produced a measurable improvement in a specific engineer's capability (not mentorship, which is knowledge transfer, but coaching, which is capability development); demonstrated comfort holding someone accountable for an outcome they did not achieve while maintaining a productive working relationship with them; demonstrated comfort navigating a people problem — a conflict, a performance gap, a motivation concern — without escalating or deferring when the situation became uncomfortable; and demonstrated genuine engagement with the work of management during shadow experiences, expressed not as willingness to help but as authentic interest in the organizational and human questions that management involves. The criteria model should specify three categories of signals: required signals (must be present for the candidate to advance to the promotion decision; these are the signals most predictive of management success), desirable signals (present in most successful managers but not in all; these inform calibration but do not gate advancement), and disqualifying signals (patterns that predict management failure regardless of IC excellence; the most common is a candidate whose approach to all difficult situations is to increase their own technical involvement rather than to influence through others). The criteria model should also specify the shadow experience requirement — the structured period during which the management candidate co-manages a small number of people alongside a current manager — because management readiness cannot be assessed without behavioral evidence, and behavioral evidence cannot be collected without a context in which the management behaviors are observable. A shadow experience of 60 to 90 days, with weekly structured reflection conversations between the candidate and the sponsoring manager, generates the behavioral data the criteria assessment requires and also functions as the first phase of the transition ramp for candidates who are subsequently promoted.

Section 2: Criteria calibration and evidence collection process. The criteria calibration process specifies how the promotion criteria are validated against management outcomes, how the evidence required for each criterion is collected, and who is responsible for collecting it. Criteria validation: each criterion should be traced to a management outcome it predicts, and the tracing should be updated annually as the company accumulates data on promoted managers' performance. A criterion that is present in the evidence record of promoted managers who subsequently performed effectively is a criterion with demonstrated predictive value; a criterion that is present in the evidence record of both effective and struggling managers is a criterion with no discriminatory value and should be removed or reweighted; a criterion that is absent from the evidence record of effective managers and present in the record of struggling managers is an inverse predictor — it predicts management struggle — and the promotion process should treat it as a disqualifying signal. Evidence collection: each criterion requires a named evidence collection method. Coaching ability is evidenced by documented 1:1 notes from the shadow experience showing development conversations rather than status updates, and by direct observation from the sponsoring manager of the candidate's behavior when a direct report described a problem the candidate could have solved for them. Accountability comfort is evidenced by a documented instance of the candidate delivering a direct performance conversation to a peer engineer — not to a junior engineer in a mentorship context, where the power dynamic reduces the discomfort, but to a peer where the social cost of directness is real. People-problem navigation is evidenced by a documented instance where the candidate identified a conflict, disengagement, or motivation concern on a project they were involved in and took a management action — a direct conversation, a design intervention, a scope change — rather than escalating to a manager or waiting for the situation to resolve. Each evidence item should be documented in a structured observation record that includes the situation, the candidate's action, and the outcome — not as a formal review artifact but as a working document that the promotion committee uses to calibrate against the criteria. Designed conversations in ChatGPT or Claude where the sponsoring manager worked through the candidate's management readiness — what situations had revealed readiness signals, what situations had revealed gaps, how similar gaps had resolved in other candidates — should be run through the open-source extractor before the promotion committee review to surface the reasoning behind the sponsoring manager's recommendation, because the recommendation made from memory at a promotion meeting omits the uncertainty and the counterevidence that were present in the original reasoning and that the committee needs to make a calibrated decision rather than a ratification.

Section 3: Promotion decision structure and calibration session protocol. The promotion decision structure specifies who has decision authority, who provides input, and what the calibration session is designed to produce. Decision authority: the direct manager's recommendation is necessary but not sufficient — a promotion decision made by the direct manager alone is vulnerable to relationship bias (the manager who has worked closely with the candidate has a strong prior toward seeing the candidate as ready, independent of the evidence) and to span of control optimization (the manager who needs to reduce their span of control has an incentive to promote a candidate regardless of readiness, because a promotion removes a direct report and is easier than restructuring the team). The promotion decision should require sign-off from the skip-level (the engineering director or VP of Engineering), with the skip-level's role being to assess whether the criteria evidence was collected systematically or was constructed post-hoc to support a decision that had already been made on relationship familiarity, and to provide a calibration check against the full cohort of engineering managers at the company rather than against the local team context. Calibration session protocol: the promotion committee review has three stages. Stage one: the sponsoring manager presents the criteria evidence for each required signal — not the conclusion ("I believe this candidate is ready") but the evidence ("in month six of the shadow experience, this candidate identified a disengagement signal in a direct report before I had noticed it, initiated a direct conversation with the direct report without prompting, and documented the conversation's outcome in their 1:1 notes within 24 hours"). Stage two: the committee tests the evidence by asking for counterexamples — situations during the shadow experience where the candidate defaulted to IC behavior under pressure, avoided a people problem, or delegated upward when the situation required a management action. If the sponsoring manager cannot provide any counterexamples, the evidence is probably not complete — all management readiness assessments over 90 days will include situations where the candidate did not perform at the management standard, and a review that presents only positive evidence is a review that has not examined the evidence systematically. Stage three: the committee makes a decision — promote, extend the shadow experience for a specific period targeting a specific evidence gap, or decline with documented reasoning that the candidate can use to understand what development would be required before a future candidacy.

Section 4: IC-to-manager transition support model. The transition support model specifies what structured support is provided to a newly promoted manager and over what timeline — the model that treats the promotion decision as the conclusion of the management development process rather than as the beginning of the management ramp phase. The ramp has four structural requirements. First, complete IC handoff: the new manager must have zero engineering deliverables — no code ownership, no active sprint tasks, no architectural decision ownership — by the end of week two. The most common early transition failure is a new manager who continues holding IC responsibilities during month one because the handoff is gradual rather than complete, producing a situation where the manager is doing two jobs partially rather than one job fully; the people management work that produces outcomes — proactive development conversations, early performance identification, upward capacity communication — requires uninterrupted attention, and it is the work that gets deprioritized when the manager is also holding IC deliverables. Second, sequenced management practice introduction: the ramp curriculum sequences the management practices that new managers develop in a specific order, because the practices build on each other and introducing them simultaneously produces overwhelm rather than competence development. The sequence is: diagnostic listening before development advising (months one to two: 1:1s are structured around understanding each direct report's current situation, goals, and blockers, with the manager in a listening posture rather than an advisory posture); development conversations before performance conversations (months two to three: structured development 1:1s distinct from status 1:1s, before the manager needs to navigate a performance situation that requires a direct conversation about inadequate work); performance conversations before delegation (months three to four: the ability to hold someone accountable for an outcome is a prerequisite for effective delegation, because delegation without accountability produces activity without outcomes). Third, weekly check-in cadence with the ramp owner: the ramp owner — the senior manager or engineering director — holds a weekly 30-minute structured conversation with the new manager for the first two months, focused on the current ramp phase rather than on team delivery. The check-in has a fixed agenda: what management situation was most challenging in the past week, what the new manager did, what they would do differently, and what specific support they need from the ramp owner in the coming week. The check-in's job is not to supervise the new manager's team — that is the new manager's job — but to provide a rapid feedback loop on the management practices the new manager is developing, because management practice errors that are caught in week three are learning; management practice errors that persist into month six are patterns. Fourth, structured first performance calibration: in month four, the new manager conducts their first direct report performance calibration with the ramp owner as a calibration partner rather than as a reviewer. The calibration produces a one-paragraph written assessment for each direct report — current performance level, development trajectory, and one specific action the manager is taking to address either a gap or a growth opportunity. The written calibration is the forcing function that converts the diagnostic listening of months one and two into an actionable management model for each direct report, and the calibration partner's role is to test whether the assessments are specific and behavioral or vague and relational — a manager whose calibration for a struggling engineer reads "they're still finding their footing" has not done the diagnostic work the ramp was designed to produce; a manager whose calibration reads "in the last three sprints, they have underestimated task complexity in the planning stage, which produces incomplete deliverables at demo time; I'm scheduled to address this directly in next week's 1:1 and will follow up in two weeks to assess whether the conversation changed the estimation approach" is a manager who has developed the diagnostic posture the ramp was designed to install.

Section 5: Manager effectiveness measurement and criteria review cadence. The effectiveness measurement model specifies what is measured for newly promoted managers, how frequently, and how the results are used to update the promotion criteria. Manager effectiveness metrics: new manager retention in role at 12 months is the primary metric — a new manager who exits the manager role within 12 months, whether through their own choice or through a performance intervention, is a transition failure; the causes of the exit (IC excellence criteria rather than management readiness criteria, absent transition support, span of control miscalibration, skip-level misalignment) are recorded in the exit documentation and used to update the criteria model. Direct report development cadence at 6 months: for each new manager, the number of direct reports with a documented development plan at month 6 divided by total direct reports is tracked; a ratio below 50% at month 6 indicates that the manager has not yet converted the diagnostic 1:1s of months one and two into actionable development conversations, which is the most common leading indicator of management struggle before the skip-level satisfaction data surfaces a concern. Skip-level satisfaction with new managers at 6 months: a structured two-question skip-level check-in for direct reports of new managers — "does your manager understand your current situation well enough to help you navigate it?" and "has your manager had a conversation with you about your development in the past 60 days?" — provides a behavioral signal on the two most critical early management practices without requiring a full engagement survey. Manager self-reported practice comfort at month four: a structured self-assessment of four practices — performance conversation, development conversation, upward capacity communication, and cross-functional coordination — on a five-point scale, with specific behavioral anchors for each scale point, produces a leading indicator of which practices are developing on track and which will require additional ramp support before they produce the downstream performance signals. Criteria review cadence: annually, the promotion committee conducts a retrospective on the prior year's promoted managers — what criteria evidence was present in the records of managers who performed effectively at 12 months, what criteria evidence was present in the records of managers who required performance interventions, and what patterns in the shadow experience documentation predict both outcomes. The criteria model is updated to increase the weight of evidence types that predict 12-month effectiveness and to add or remove required signals based on the retrospective's findings. AI-assisted design conversations from the original criteria development sessions — the sessions where the VP of Engineering or senior manager worked through what management readiness looks like at this specific company, what has distinguished effective from struggling managers in the past, and what shadow experience structure would generate the evidence the criteria require — should be run through the open-source extractor before each annual criteria review to recover the original reasoning behind each criterion, because criteria that were added in response to a specific historical failure (a promotion that went wrong in a specific, memorable way) are the criteria most vulnerable to being removed by a new committee member who sees them as overly specific and who lacks the context of the failure that motivated them; the extracted reasoning surfaces the motivation behind the criterion and allows the committee to evaluate whether the motivation is still load-bearing rather than whether the criterion still sounds right in the abstract.

FAQ

Further reading

• Engineering career ladder decision record — the career ladder model and the promotion criteria model are structurally coupled: the ladder determines what promotion paths exist and what evidence is collected in each path's performance review process, and the promotion criteria model is constrained by whatever assessment infrastructure the ladder provides; a ladder with no explicit IC-to-manager branch — where the management track is treated as a higher rung on the same ladder rather than as a structurally distinct role with structurally distinct criteria — will produce promotion decisions that inherit the ladder's IC assessment framework by default, because there is no alternative framework defined; building explicit management promotion criteria requires first building an explicit management track on the career ladder with a branch point at the IC-to-manager transition that signals to everyone in the organization that management readiness and IC excellence are different properties requiring different evidence.
• Engineering onboarding ramp decision record — the IC onboarding ramp problem and the new manager transition problem share the same structural failure mode: the hire or promotion decision is treated as the conclusion of a selection process rather than the beginning of a structured capability transfer; in both cases, the ramp owner is the senior person responsible for the new hire's or new manager's effectiveness, and in both cases the most common failure is treating the ramp as an orientation process — making resources available, scheduling introductory meetings, providing documentation — rather than as a proactive capability transfer process where the ramp owner actively manages what the new person is doing and experiencing in each phase; the IC onboarding ramp and the new manager transition ramp differ in content but share the same design principles: complete handoff of prior responsibilities before the ramp begins, sequenced practice introduction that builds on each phase, regular check-ins between the ramp owner and the person ramping that focus on the ramp phase rather than on output metrics, and a defined endpoint that specifies what the person must be doing independently before the ramp formally concludes.
• Engineering manager span of control decision record — the span of control model and the promotion criteria model interact on two dimensions: first, the span assigned to a new manager should be calibrated against the new manager's transition phase rather than against the healthy steady-state span for an experienced manager, because the management work that is most demanding in the transition period — diagnostic 1:1s for each direct report, development conversations, early performance identification — scales with the number of direct reports and a new manager who is still developing these practices will be overwhelmed by a full span before they have the judgment to execute the practices efficiently; second, the span of control that produced a reluctant manager's adequate-but-not-good management performance is not a span of control problem but a criteria problem — a manager who is managing five engineers adequately but not developing any of them is not a manager who needs a smaller team, but a manager whose promotion criteria selected for IC excellence rather than for development motivation, and reducing the span does not change the motivation model that is producing the absence of development behavior.
• Engineering hiring bar decision record — the internal promotion criteria model and the external management hiring bar are complements in the management pipeline: the internal criteria determine what the company grows internally, and the external hiring bar determines what the company imports; a company that promotes poorly from inside — selecting on IC excellence and experiencing high transition failure rates — will increasingly rely on external management hiring to fill the management capacity that internal promotions are not reliably producing; the external management hiring bar must then be calibrated against a management team that was built partially through failed internal promotions, which is a harder calibration problem than calibrating an external bar against a management team built through deliberate, well-documented criteria; the internal promotion criteria model is therefore an input to the external hiring bar's calibration, and a company that improves its internal criteria will find that its external management hiring becomes easier to calibrate because the internal management team's effectiveness provides a clearer and more defensible bar to hire against.
• Open-source extractor — the management readiness assessment conversations that produce the most accurate criteria are typically the design sessions where the VP of Engineering or senior manager worked through specific historical promotion decisions with ChatGPT or Claude: what the candidate's evidence looked like, what the committee's uncertainties were, what counterexamples were raised and how they were resolved, what happened after the promotion and whether the prediction turned out to be correct; running those sessions through the extractor before the annual criteria review recovers the explicit reasoning that was present in the original assessment conversation and that is absent from the formal promotion record, which typically documents only the conclusion rather than the evidence evaluation that produced it; the extracted reasoning surfaces the specific behavioral observations that were considered predictive at the time of each promotion, which is the ground truth the annual criteria review needs to evaluate whether the criteria are selecting for the properties that actually predict management success at this company rather than the properties that look like management readiness in the abstract.