Build an Engineering Team People Want to Come Back To  🔗 **Source:** [Build an Engineering Team People Want to Come Back To](https://leaddev.com/culture/build-an-engineering-team-people-want-to-come-back-to) > [!note]+ Document Notes > tags: [engineering-management, boomerang-employees, team-culture, career-feedback, offboarding, contributor-retention] > author: Meg Adams > source: LeadDev > reference: https://leaddev.com/culture/build-an-engineering-team-people-want-to-come-back-to > date created: 2026-09-24 11:31 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: Build an engineering team people want to come back to > <!-- deep-read:sections start --> ## 🧭 Table of contents - [📝 Summary](#-summary) - [📖 Main content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning styles assessment](#-learning-styles-assessment) - [📊 Accessibility assessment](#-accessibility-assessment) - [🌍 DEIB assessment](#-deib-assessment) - [✅ Logical validity](#-logical-validity) - [🛡️ DARVO analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🔄 Refactor appendix](#-refactor-appendix--build-an-engineering-team-people-want-to-come-back-to) ## 📝 Summary - **What happened.** LeadDev published a 1,444-word essay by Meg Adams on 2026-09-24 in its Culture section, under nine headings plus a three-item "Key takeaways" box. It opens with one anecdote: a former report left, worked elsewhere, and asked to return, describing the other team as "grey" with everyone "in their own lane." - **The claim.** Managers can't always control pay, but they can design a team people want to return to. Three levers, stated at ¶9: a team that exists apart from its current members, feedback sized to a career rather than a review cycle, and a resignation treated as a win. - **The mechanism.** In a seasonal retail store with fixed pay and eight-week stints (¶6 to ¶8), retention was impossible, so return became the goal: the same students back each June and December, already trained. Adams reports applying that to every engineering team since "to the same result" (¶9). No number backs "same result." - **Why it matters.** For DevRel programs and open-source contributor pipelines, churn is the default state. Contributors leave for jobs, kids, and burnout, and many come back. An essay that makes return rather than retention the design target maps onto that world more directly than most management writing. - **Key stakeholders.** Meg Adams (author; the byline names no employer). LeadDev (publisher; all 10 anchors point back to its own catalogue). Daniel Terhorst-North, named at ¶13 and linked through a LeadDev article. Engineering managers, and by extension community and DevRel leads. - **Risks and tradeoffs.** The evidence is one returner's sentence plus "several times now, on a few different teams" (¶4). Adams flags the survivorship problem herself at ¶21. No study, headcount, or return rate appears. This analysis marks each practice as either a stated mechanism or a story taken on trust. - **Bottom line.** Two ideas hold regardless of the thin evidence: describe a team by what will still be true two years after everyone has left (¶10), and treat the first minute after a resignation as the only part anyone remembers (¶18). Both convert directly into contributor-program practice. ## 📖 Main content **The core argument.** Adams opens with a signal. A former report came back and described the interim employer as "grey – with everyone in their own lane" (¶1). ¶3 argues the returner is a weighted signal: the engineers who leave, want to return, and are wanted back have the most options, so their return is a comparison made against real alternatives. ¶4 generalizes from "several times now" to an operating assumption that she'll work with her people for many more years. Everything after that is what follows from taking a 20-year horizon seriously. **Where the model comes from.** ¶5 to ¶9: seasonal retail, college-student staff on for a handful of weeks, pay fixed and identical at every store on the block (¶8). Retention was impossible, so Adams optimized for return, which cut from-scratch onboardings each season. ¶9 claims the same result on engineering teams and names the three pillars. **Pillar one: the team exists apart from its members.** ¶10 asks what would still be true two years after everyone has gone and answers: the standard of performance, the problems, the teamwork, the memories. The next four headings break those out. Protect the bar (¶11 to ¶12) is a causal chain ending with the strongest engineers, who absorb the most work and have the most options. Real choice (¶13) is a practice: stack rank projects or teams, with Terhorst-North cited. Work together by design (¶14) rests on the boomerang alone. Make memories (¶15) appeals to "the science of memory" without naming a source. **Pillar two: feedback sized to a career.** ¶16: review cycles buy specificity but shorten the horizon. Stretching it lets a manager name a habit that works now and will block the person two levels later, while it's still a small correction. ¶17 adds the self-interested version: if she'll share a room with this person in 5, 8, or 20 years, she'd rather they'd already grown. **Pillar three: a resignation isn't a loss to manage.** ¶18 is the sharpest paragraph: the resignation moment is when the person learns what you think of them, and backfill, transition, and coverage all have to happen, but not first, because the first reaction is the only part remembered. ¶19 makes the analogy: a team that celebrates people leveling up already knows how to react, and a resignation is the same event. Her script is to close the laptop, lean in, and ask for everything. ¶20 adds a calendar cadence of check-ins after departure, imperfectly kept. **The self-critique and the bet.** ¶21: returners are by definition the ones the team worked for, a biased sample and the only one that reaches her. More honest than the genre usually manages, and it undercuts the "same result" claim at ¶9, since the result is measured by the sample that returned. ¶22 closes with an asymmetric wager: assume 20 years and be wrong, lose nothing; assume two years and be wrong, lose every skipped conversation. Sound as decision theory, with one gap: cadences and early hard conversations cost manager hours, so "lost nothing" overstates it. **Anecdote or mechanism: the ledger.** The reader context asks which practices rest on the boomerang story and which on a stated mechanism. | Practice | Heading | Rests on | Where | |---|---|---|---| | Returners are a weighted culture signal | intro | Revealed-preference reasoning plus one anecdote | ¶1 to ¶4 | | Optimize for return when retention is impossible | Where I learned it | Retail natural experiment (fixed pay, fixed stint) | ¶5 to ¶9 | | Describe the team by what survives turnover | The team has to exist apart… | Stated mechanism (two-year test) | ¶10 | | Hold the performance bar | Protect the bar | Stated causal chain, no data | ¶11 to ¶12 | | Stack-rank projects with the person | Give people real choice | Practice described; "just good business" asserted; Terhorst-North cited | ¶13 | | Design collaborative projects | Design opportunities to work together | The boomerang anecdote alone | ¶14 | | Manufacture novelty at milestones | Make memories | "Science of memory" gestured at, uncited | ¶15 | | Feedback against a career horizon | Feedback sized to a career | Stated mechanism plus self-interest argument | ¶16 to ¶17 | | Celebrate resignations first, process second | A resignation isn't a loss | Stated mechanism plus author's script | ¶18 to ¶20 | | Returners are a biased sample | Final thoughts | Author's own caveat | ¶21 | | Long horizon is cheap to be wrong about | Final thoughts | Asymmetric-payoff argument | ¶22 | ## 🗒️ Notes #### Before the first heading (Key takeaways and the boomerang) - [L0] to [L2]: returners are a strong culture signal, especially from people with alternatives; build a team that outlasts its roster; react to resignations like wins. - ¶1 to ¶2: a former report returns and describes the interim team as "grey – with everyone in their own lane." Adams reads the implication for her own team and declines to judge the other company. - ¶3: the signal is weighted because the people who leave, want back, and are wanted back have the most options. ¶4: "several times now, on a few different teams," no count. > [!quote] [L0] > A returning employee is a strong culture signal, especially from someone with real alternatives. > [!quote] [¶3] > This means a return is a comparison someone made with real alternatives available to them. #### Where I learned it - ¶5 to ¶7: a large seasonal retail store; college students on for a handful of weeks; near-zero training window. Retention wasn't the objective because it wasn't possible. The controllable outcome was the same people back in June and December, so each season had fewer brand-new employees. - ¶8: pay was set and identical across every store on the block, so the experience was the only variable. The essay's cleanest reasoning: a natural experiment with pay held constant. - ¶9: applied "within all of my engineering teams since, to the same result." No result is named. The three pillars appear here. > [!quote] [¶8] > The only variable I controlled was whether the eight weeks had been worth repeating. > [!quote] [¶9] > a team that exists apart from the people in it, feedback sized to a career, and a resignation you don’t treat as a loss. #### The team has to exist apart from the people currently in it - ¶10: a roster answer to "what's the team like" has a short shelf life. The durable answer is what would still be accurate two years after everyone has gone: performance standard, problems, teamwork, memories. The two-year test is the reusable tool. > [!quote] [¶10] > Ask someone what a team is like and they’ll usually describe who’s on it. That’s a fair answer, but it has a short shelf life. #### Protect the bar - ¶11: performance problems are visible in team dynamics before they reach the manager, because teammates cover the work and absorb the blast radius. - ¶12: handling a case supportively and efficiently affirms the standard is required. The strongest engineers feel this first, because they absorb the work and have somewhere else to go. A stated causal chain, no data. > [!quote] [¶12] > it affirms that a standard of performance isn’t decorative, it’s required. #### Give people real choice about what they work on and make sure they're growing - ¶13: "What do you want to learn this year" gets asked once and dropped. Adams has people stack rank the projects or teams they want and uses the ranking in assignment. Cites Daniel Terhorst-North; the link goes to a LeadDev article about his ideas. Rests on practice plus assertion, no outcome metric. > [!quote] [¶13] > making actual assignment decisions with the person and where their interest and growth is a genuine input is just good business. #### Design opportunities to work together - ¶14: the returner's complaint was solitude, not dullness. Prescription: design projects so people learn from and teach each other. The only heading whose evidence is entirely the ¶1 sentence. > [!quote] [¶14] > It wasn’t that the work was dull, but that it was solitary. #### Make memories - ¶15: novelty is the quickest path to a shared memory; milestones and celebrations are the lowest-lift venue for being "a little zany" (themed parties, unexpected excursions). "The science of memory is incredibly interesting," with no author, book, or study named. #### Feedback sized to a career, not a review cycle - ¶16: review loops serve the current cycle. Stretching the horizon changes the feedback's shape: the hard thing gets said earlier, about a habit that works now and will hold the person back two levels later. - ¶17: the self-interested framing (5, 8, or 20 years in the same room). Rests on a mechanism plus a self-interest argument; both are reasoning, both testable. > [!quote] [¶16] > It can help me say the hard thing earlier, when it’s still a small correction. > [!quote] [¶17] > Dodging a hard conversation now just means I’ll have a harder conversation later. #### A resignation isn't a loss to manage - ¶18: the resignation moment reveals what you think of the person. The machinery (backfill, transition, coverage) has to happen, but not first, because the first thing is the only part remembered. - ¶19: the level-up analogy. Anything less than enthusiasm reveals how conditional the interest was. Script: close the laptop, lean in, "how exciting! Tell me literally everything." - ¶20: a calendar check-in cadence after departure (meeting, email, or text), admittedly inconsistent, which she says clears "a pretty low industry bar." No source for the bar. > [!quote] [¶18] > The moment someone tells you they’re leaving is the moment they find out what you actually think of them. > [!quote] [¶19] > if my reaction is anything less than enthused, I’ve communicated exactly how conditional my interest in them always was. > [!quote] [¶20] > Effort is the difference between being a manager someone used to have and someone who was in their corner either way. #### Final thoughts on building engineering teams people want to come back to - ¶21: the survivorship caveat. Returners are the ones it worked for; the person who found the team too chatty or didn't feel included never writes back. - ¶22: compensation isn't always controllable; team design is. The asymmetric bet: 20-year colleague and wrong, lose nothing; two-year hire and wrong, lose years of conversations. [L2] restates the third pillar. > [!quote] [¶21] > They’re a biased sample and they’re also the only sample that talks to me. > [!quote] [¶22] > Treating the horizon of our working relationship as long is cheap to be wrong about. > [!quote] [L2] > React to resignations like wins, not losses. ## 📚 References The article contains 10 anchors (from `leaddev-team_links.txt`): two self-links (an empty anchor and the hero image, text "article hero – leadership") and eight LeadDev articles. Every citation is to the publisher's own catalogue: no external source, no study, and no link to Daniel Terhorst-North's own writing. - [Build an engineering team people want to come back to](https://leaddev.com/culture/build-an-engineering-team-people-want-to-come-back-to): the source. LeadDev, Culture, Meg Adams, 2026-09-24, 1,444 words. - [3 ways to bounce back from sudden team cuts](https://leaddev.com/hiring/3-ways-to-bounce-back-from-sudden-team-cuts): anchored on "someone leaving" at ¶3. Context link; supports nothing in the signal argument. - [Training innovation: the key to long-term success](https://leaddev.com/culture/training-innovation-the-key-to-long-term-success): anchored on "train" at ¶6. Editorial cross-link, not evidence. - [4 rituals that revolutionized my engineering teams](https://leaddev.com/culture/4-rituals-that-revolutionized-my-engineering-teams): anchored on "engineering teams" at ¶9, where the "same result" claim sits. Adjacent to Make memories. - [Navigating engineering performance reviews](https://leaddev.com/hiring/navigating-engineering-performance-reviews): anchored on "handle difficult performance" at ¶12. The nearest thing to support for Protect the bar, and another opinion piece. - [Shaped by demand: the power of fluid teams](https://leaddev.com/management/shaped-by-demand-the-power-of-fluid-teams): anchored on "Daniel Terhorst-North's thinking" at ¶13. A LeadDev article about his ideas. Primary source, not linked in the article: [Daniel Terhorst-North's site](https://dannorth.net/). - [Provide growth opportunities for your team this summer](https://leaddev.com/management/provide-growth-opportunities-your-team-summer): anchored on "Design projects" at ¶14. Supports the collaborative-design prescription by adjacency only. - [Staff engineers must rebuild team culture](https://leaddev.com/leadership/staff-engineers-must-rebuild-team-culture): anchored on "Creating an environment" at ¶15. - [How to create an effective offboarding plan when you're leaving your job](https://leaddev.com/hiring/how-create-effective-offboarding-plan-when-youre-leaving-your-job): anchored on "someone tells you they're leaving" at ¶18. Written from the leaver's side, so it complements rather than supports the manager-reaction argument. ## 🗣️ Paraphrase Adams argues that the strongest evidence a team culture is good comes from people who left, saw an alternative, and asked to come back; they had options, so their return is a real comparison. She learned to design for return rather than retention in a seasonal retail job where pay was fixed and every hire was gone in eight weeks, so the only variable she owned was whether the stint had been worth repeating. Three practices follow. Define the team by what survives turnover: a held performance standard, interesting problems, collaborative project design, shared memories. Give feedback against the length of a career instead of a review cycle, so the small correction happens early. Treat a resignation like a promotion: enthusiasm first, backfill second, a check-in cadence afterward, because the first reaction is what gets remembered. She closes by admitting the returners are a biased sample, then makes the bet explicit: assuming a 20-year relationship costs almost nothing if wrong; assuming a two-year one and being wrong costs every conversation you skipped. ## 📰 Newsletter **Contributors leave. Build for the ones who come back.** LeadDev ran a short essay this week by engineering manager Meg Adams that reads like it was written for open-source maintainers. In a seasonal retail job with fixed pay and eight-week stints, retention was impossible, so she optimized for return: the same students back every June and December, already trained. She's run engineering teams on the same idea since. Three moves. Describe your team by what would still be true two years after everyone left: the bar you hold, the problems, how people collaborate, what they remember. Give feedback sized to a career, not a release cycle. And when someone says they're leaving, celebrate before you open the backfill ticket, because the first reaction is the only part they'll remember. Swap "engineer" for "contributor" and this is a WordPress team-rep playbook. Core, Docs, Polyglots, and Training all run on people who cycle in and out. Her caveat applies to us too: the people who come back are the ones it worked for. Worth 6 minutes and one question at your next contributor meeting: what about this team survives all of us leaving? [Read it on LeadDev](https://leaddev.com/culture/build-an-engineering-team-people-want-to-come-back-to). Every link inside points to another LeadDev article; there's no outside research behind it. ## ✅ Action items #### DevRel - Developer advocacy: write the two-year test (¶10) for the DevRel program itself: what a developer would still describe accurately after every current advocate has moved on. Anything on the list that's a person's name is a single point of failure. - Education: add a returning-contributor track beside the new-contributor track. A returner needs a delta ("what changed since you left"), not full onboarding; ¶7's savings only appear if you don't re-onboard from scratch. - Developer experience: audit quickstarts for ¶14. A sample one person completes alone teaches the tool; one that expects a pair or a review teaches the community. Ship one of the second kind. - Developer marketing: track boomerang developers (stopped using the platform, came back) as a named segment. ¶3 calls that the strongest culture signal; report it with the ¶21 caveat beside it. - Community management: script the first response to "I'm stepping back" in every channel, as Adams scripts hers at ¶19. Enthusiasm first, process second. - Internal/external/cross-project: set a quarterly alumni check-in cadence for former ambassadors and speakers (¶20), tracked, and measure 12-month re-engagement against the pre-cadence baseline. #### OSPO - Policy and governance: write down whether former employees who contributed to Pantheon's open-source repos keep maintainer rights after departure, and for how long. ¶10 argues for a role-based answer, not a person-based one. - License compliance: no direct action from this source; nearest relevance is CLA or DCO continuity when a contributor returns under a new employer. - Security management: alumni access is an attack surface. Run the ¶20 cadence and the access review on the same calendar. - Software inventory: no direct action from this source. - Developer support: fold career-sized feedback (¶16) into maintainer mentoring guidance. A reviewer who only comments on this PR misses the habit that blocks the contributor's path to maintainer. - Community engagement: ask funded upstream projects the ¶10 question at MoU review. A one-person project is a funding risk regardless of its code. - Strategic alignment: add contributor return rate (inactive 6+ months, then active) as a KPI beside new-contributor count. #### Community - Maintainers: run a governance clinic on the ¶18 hour. Draft the first message to a departing co-maintainer before you need it; the release handoff is the second message. - Contributors: build a re-onboarding path that assumes prior context: a changelog of governance, tooling, and roadmap changes since their last commit. Measure returner time to first merged PR against new-contributor time. - Contributors: adopt ¶13 literally. Let active contributors stack rank the components they want each cycle, use it in assignment, and report the share who got a top-two choice. - Sponsors: when a sponsored maintainer leaves, publish what the sponsorship funded that survives them (docs, tests, automation), which is ¶10 applied to a budget line. - Ecosystem: someone who left WordPress for Drupal and came back holds the comparison ¶3 calls the signal. Ask them. - On the caveat: before citing any "people love coming back" story, pair it with exit data from people who didn't return (¶21), or don't cite it. #### WordPress - Core: no direct action from this source; nearest relevance is asking the ¶10 question of each component: what survives its current maintainer leaving? - Core Performance: no direct action from this source. - Core AI: no direct action from this source. - **Core Program Team:** release squads rotate people. Adopt the ¶18 rule for squad departures: first message is thanks, second is the handoff checklist. Draft both templates. - Mobile: no direct action from this source. - Playground: no direct action from this source. - Community: WordCamp and meetup organizers cycle out constantly. Build a returning-organizer path (delta briefing, not full training) and track how many past organizers return within two years. ¶7's frame fits organizer teams best. - Training: a short lesson for team reps on the three pillars, with this analysis's anecdote-versus-mechanism ledger attached. - Marketing: no direct action from this source; nearest relevance is sourcing contributor stories from returners with the ¶21 bias stated in the copy. - Support: forum volunteers step back and return. Script the first response (¶19) and keep a light cadence (¶20) for high-volume alumni. - Design: no direct action from this source. - Documentation: "what changed since you last contributed" pages per team, refreshed each major release. The cheapest re-onboarding artifact. - Photos: no direct action from this source. - TV: no direct action from this source. - Polyglots: compare returning translators' time to first approved string against new translators, per locale. If ¶7 holds, GTE effort should go toward inviting returners back. - Accessibility: no direct action from this source. - Meta: a profile badge for "returned after 6+ months inactive" would make the ¶3 signal visible with no new infrastructure. - CLI: no direct action from this source. - Hosting: no direct action from this source. - Tide: no direct action from this source. - Openverse: no direct action from this source. - Themes: review volunteers churn; apply the Support script and cadence. - Plugins: same as Themes, plus ¶12: the strongest reviewers absorb the slow reviewers' backlog and leave first if the bar isn't held. #### The WPOCC - Programs: The WP Open Community Collective funds people, and people leave. Design fellowships and stipends so the artifact outlives the fellow (¶10): a documented process, a test suite, or a governance template, not only hours. - Programs: add a quarterly alumni cadence (¶20) for two years after each fellowship ends, tracked, so The WPOCC learns which alumni return. - Projects: use the ¶18 rule as a governance-clinic case: how a project responds in the first hour after a maintainer resigns. Collect three real examples. - Funding experiments: fund two contributor cohorts on identical terms; give one a scripted celebratory offboarding plus quarterly check-ins, the other the default. Measure re-engagement at 12 months and publish both numbers. ### FAIR - Technical independence: no direct action from this source; nearest relevance is that a mirror maintained by one person fails the ¶10 test regardless of code quality. - Policies and covenants: write a maintainer succession clause into operator agreements, so the process exists before the departure. ¶18 only works if the process is already written. - Productized governance artifacts: add a "survives the current maintainers" row to operator scorecards: are documentation, keys, release automation, and decision rights role-bound or person-bound? ## 🧪 Hypotheses 1. Contributors who receive a scripted celebratory first response to a step-back announcement within 24 hours return to active contribution within 12 months at a rate at least 10 percentage points higher than contributors who receive a process-first or no response. 2. Returning contributors given a delta briefing reach first merged PR at least 40% faster than first-time contributors on the same project. This is ¶7's retail claim transferred to code. 3. On teams carrying an unaddressed low performer for 6+ months, regretted attrition among top-quartile contributors runs at least twice the team's baseline, as ¶12 predicts. 4. Contributors who stack rank components and receive a first or second choice show 12-month retention at least 15 percentage points higher than contributors assigned without ranking (¶13). 5. Returner-only data overstates satisfaction: non-returners from the same cohort score at least one dimension (inclusion, communication load, project fit) a full point lower on a 5-point scale than returners do. This tests ¶21. 6. In alumni interviews 6+ months after departure, the manager's initial reaction is named unprompted more often than any element of the transition process (¶18). ## 📊 Methodology - **H1.** Two community programs with comparable churn. One adopts a scripted first response to step-backs within 24 hours; the other changes nothing. Track every announced step-back for 6 months; measure any commit, forum answer, or event role at 12 months. Pre-register the 10-point threshold. Log all contact so extra check-ins don't confound the script. - **H2.** From activity logs on three repositories, find contributors with a 6+ month gap who returned; match each to a first-timer who started the same month. Compare days to first merged PR, and within returners, compare those who got a delta briefing against those who didn't. - **H3.** With HR records: top-quartile voluntary attrition on teams carrying an unresolved case 6+ months, against baseline. Open-source proxy: contributors reviewing more than 2x their share of another's PRs, and whether they go inactive within 6 months. - **H4.** Run the ¶13 ranking in one cohort for one release cycle. Compare 12-month retention by placement. Control the selection effect by comparing rankers who got a low choice against rankers who got a high one, not rankers against non-rankers. - **H5.** Survey returners and non-returners from one alumni list with the same instrument. The gap between means is the finding; its size is how much to discount returner testimonials. - **H6.** Semi-structured interviews with 20 alumni, 6 to 24 months out, opening with "tell me about the day you gave notice." Code what gets named first. - **Controls throughout.** Define "return" before collecting. Separate voluntary departures from layoffs, which ¶3 doesn't cover. Report counts, not percentages alone; in small programs a 10-point shift can be two people. ## ❓ Questions - What is the actual return rate on Adams' teams, over what period? "Several times" (¶4) could be 3 or 30. - ¶9's "same result" in engineering: fewer new hires per year, higher return count, shorter onboarding, or something else? - Does the ¶3 argument survive layoffs? A rehired laid-off engineer made no culture comparison. The essay covers voluntary departures only. - What does the ¶20 cadence look like: how often, for how long, does it taper? - Which memory finding does ¶15 mean? The leap from a themed party to a returning engineer is unmeasured. - ¶22 says being wrong about a 20-year horizon costs nothing. What do the cadence, the early hard conversations, and the ranking exercise cost in manager hours per report per year? - Has Adams run an exit survey or a non-returner interview (¶21), and did it change any of the three pillars? - In contributor programs with no manager and no 1:1, who owns the ¶18 first reaction: a maintainer, a team rep, a bot? ## 🔮 Future research - **Return rate as a community health metric.** CHAOSS tracks new, active, and inactive contributors but not returners. A defined metric (inactive 6+ months, then active) with cross-project baselines would let ¶7 be tested at scale. - **The first-reaction effect.** ¶18 resembles peak-end effects in the memory literature but hasn't been studied for workplace departures. H6 is a first step. - **Career-horizon feedback and its reception.** Nobody has measured whether recipients act on two-levels-ahead feedback (¶16) more than on cycle-bound feedback, or whether it lands as presumptuous. - **Survivorship correction for culture testimonials.** ¶21 names the bias; the genre never corrects for it. A method that pairs returner testimony with non-returner exit data would raise the evidentiary floor. - **Role-bound versus person-bound assets.** The ¶10 test implies a scorecard: for each capability a project depends on, is it held by a role or a person? Validate an audit instrument against real maintainer transitions. - **Solitude and departure.** ¶14 rests on one word, "solitary." Test whether review-network density and pairing predict attrition. ## 💡 Implications - **Practice.** For DevRel and contributor programs the design target shifts from retention (not controlled) to return (influenced). What you build changes: delta briefings, scripted first responses, an alumni cadence. None needs budget; all need an owner. - **Policy.** Access policies should assume return. A contributor who loses maintainer rights on departure and re-earns them from zero is a two-year hire in ¶22's terms. Time-boxed retention of rights, with a security review, treats the relationship as long. - **Security.** ¶20's relationship posture and access hygiene pull in opposite directions. Run the check-in cadence and the access review on the same calendar. - **Education.** The anecdote-versus-mechanism ledger is itself a teaching artifact: teaching team reps to separate story from argument before adopting a practice is a reusable skill. - **Funding models.** Fellowships should fund artifacts that survive the fellow (¶10). A funder that can't name what a stipend left behind funded hours, not capacity. The WPOCC and FAIR can write this into grant terms. - **Evidence standards.** A LeadDev essay linking eight LeadDev essays is a closed loop. The ideas may be right; the links won't tell you. ## 🧠 Learning styles assessment - **Reading level estimate.** Flesch-Kincaid Grade Level on the article body (1,375 words after stripping paragraph markers, 62 sentences, 22.2 words per sentence, 1.48 syllables per word): grade 10.5. Flesch Reading Ease: 59.1. Vocabulary is plain; sentence length raises the grade, with several sentences past 40 words (¶3, ¶11, ¶13). #### 👀 Visual learners - Present: one hero image with the generic anchor text "article hero – leadership," and a Key takeaways box. No diagrams or tables. - Improve: the three pillars at ¶9 and their sub-headings are a tree; draw it. The ¶22 bet is a two-by-two (horizon long or short, right or wrong) and belongs in a table. #### 🔊 Aural learners - Present: none indicated in the captured page. - Improve: the ¶19 script is a spoken artifact and would land better heard. A 3-minute audio summary would serve this group. #### 🗣️ Verbal learners - Present: strong. Named concepts ("feedback sized to a career"), an analogy (resignation as level-up, ¶19), a memorable image ("grey – with everyone in their own lane," ¶1). - Improve: "fluid teams" (¶13) and "blast radius" (¶11) go undefined; "the science of memory" (¶15) needs a source or a cut. #### 🧩 Kinesthetic learners - Present: two repeatable actions, the ranking exercise (¶13) and the post-departure cadence (¶20), plus a rehearsable script (¶19). - Improve: no checklist or template. A "first hour after a resignation" checklist and a sample check-in message would fix it. #### 🧠 Logical learners - Present: the ¶8 controlled variable, the ¶12 causal chain, the ¶22 payoff asymmetry, and the ¶21 self-critique are explicit reasoning. - Improve: no numbers anywhere. One return count and one time window would satisfy this group. #### 🧑‍🤝‍🧑 Social learners - Present: the essay is about relationships and includes dialogue-shaped moments (¶1, ¶19). LeadDev pages carry comments. - Improve: no discussion prompt. ¶10's two-year question is a ready-made one for a team meeting. #### 🧘‍♀️ Solitary learners - Present: the ¶10 test and the ¶17 self-interest framing invite private reflection; the essay is short enough to read twice. - Improve: no explicit prompt. "Write down your reaction the last time someone resigned" would turn ¶18 into an exercise. #### 🌿 Naturalistic learners - Present: the seasonal retail metaphor (June and December, returning each season) is a cycle metaphor. - Improve: frame contributor cycles as an ecosystem: departure as dispersal, return as re-seeding, the team as habitat. ¶10 already says this without the words. ## 📊 Accessibility assessment Judged from the captured text, heading list, and link list of the leaddev.com page. Properties not recoverable from the capture are marked unverifiable. #### Screen reader semantics - Present: nine headings in logical order; a Key takeaways list up front; mostly descriptive link text ("Daniel Terhorst-North's thinking"). - Improve: the hero image's text "article hero – leadership" is a generic label, and the image self-links to the page. WCAG 2.2 AA 1.1.1 (Non-text Content) and 2.4.4 (Link Purpose in Context). One anchor has empty text: 2.4.4 and 4.1.2 (Name, Role, Value) if it's focusable. #### Keyboard navigation - Present: a standard article page with no custom controls indicated. - Improve: unverifiable. Confirm visible focus (2.4.7) and that the empty-text anchor isn't an unnamed focus stop. #### Contrast and readability - Present: unverifiable from the capture. - Improve: verify body text at 4.5:1 (1.4.3) and the takeaways box if tinted. A 22-word average sentence is a readability cost separate from WCAG. #### Text resizing and reflow - Present: a responsive template; no text baked into images beyond the hero. - Improve: verify reflow at 320 CSS pixels (1.4.10) and 200% zoom (1.4.4) for the takeaways box and sidebar promotions. #### Media and captions - Present: no audio or video in the captured page; 1.2.x doesn't apply. - Improve: if an audio version is added, provide a transcript (1.2.1) and don't autoplay. #### Forms and interactions - Present: newsletter signup and comments are the only interactive elements, outside the article content. - Improve: verify labels (3.3.2), error identification (3.3.1), and 24 by 24 CSS pixel targets (2.5.8) on share buttons. #### Standards reference - Gaps from the capture: **1.1.1** (generic hero text), **2.4.4** (self-linking hero and an empty anchor), possibly **4.1.2**. The content itself is text-only prose with good heading structure, which is the easy case. ## 🌍 DEIB assessment #### Cultural breadth - Present: one work context (US seasonal retail with college-student staff, June and December peaks) and one industry. - Improve: one example outside the US and outside employment, such as an open-source project, where "fixed pay" becomes "no pay." #### Inclusive language - Present: gender-neutral throughout ("someone," "they"). Idioms ("in their own lane," "in their corner") are explained by context. - Improve: "blast radius" (¶11) is ordnance jargon; a plainer phrase would serve non-native readers. #### Neurodiversity support - Present: headings every 1 to 3 paragraphs, a takeaways box, a consistent claim-then-reasoning pattern. - Improve: ¶15 treats novelty ("unexpected excursions") as universally welcome; for autistic and anxious team members it's a cost. ¶21 half-acknowledges this ("chattier than they liked") without connecting it back. Add "with opt-outs" to Make memories. #### Representation - Present: one author, one unnamed returner, one named outside thinker. - Improve: no voice from someone who left and didn't return, which ¶21 admits is the missing sample. #### Psychological safety - Present: warm and non-judgmental toward leavers; ¶2 declines to judge the other company; ¶21 models admitting evidentiary limits. - Improve: Protect the bar (¶11 to ¶12) is told from the manager's and strong performers' side. A line on what "supportively" means for the person on the receiving end would keep it from reading as managing people out. ## ✅ Logical validity #### Strong reasoning - **Selection-based inference, ¶3.** People who leave, want to return, and are wanted back have the most options, so their return is a comparison against real alternatives. Correct revealed-preference reasoning; clarify by stating the denominator. - **Controlled variable, ¶8.** Pay was fixed across every competing employer, so experience was the only thing Adams could vary. The essay's closest thing to a natural experiment, and what makes the lesson transferable. - **Causal chain, ¶11 to ¶12.** Teammates absorb the work, so they see the problem first; holding the bar signals the standard is real; the strongest, who absorb most and can leave most easily, respond first. Strengthen by defining "supportively and efficiently" in time and steps. - **Analogical, ¶19.** If the team celebrates people leveling up, and a resignation is another instance, the same reaction applies. Holds for a step up; weaker for lateral or forced moves. - **Asymmetric payoff, ¶22.** Assume long and be wrong: small loss. Assume short and be wrong: large loss. Valid as decision under uncertainty. Clarify by pricing the small loss instead of calling it nothing. - **Self-limiting disclosure, ¶21.** Naming the survivorship bias in one's own evidence is a concession against interest and raises the credibility of the rest. #### Fallacies - **Hasty generalization, ¶4 and ¶9.** "Several times" and "to the same result" generalize from an uncounted set of returners to a rule. *Fix:* give the count, window, and team sizes. - **Survivorship bias, named at ¶21 but not corrected.** The ¶9 claim still stands on the returner sample. Naming a bias doesn't remove it. *Fix:* pair with non-returner exit data, or soften ¶9 to "the returners I hear from say so." - **Appeal to unnamed authority, ¶15.** "The science of memory" is invoked without a citation, so the reader can't check the link from a themed party to a returning engineer. *Fix:* one named source, or drop the appeal. - **Unsupported quantification, ¶20.** "a pretty low industry bar" has no source. *Fix:* cite a survey on manager-alumni contact or say "in my experience." - **Understated cost, ¶22.** "I've lost nothing" ignores the manager hours the essay's own practices consume. *Fix:* estimate them; the argument survives at "lost little." - **Single-case support, ¶14.** Design opportunities to work together rests on one returner's word "solitary." *Fix:* present it as a hypothesis from one case. - **Closed citation loop, throughout.** All eight outbound links go to LeadDev. Not a reasoning error, but following them yields more opinion, not evidence. #### Tone of voice - First person, reflective, plain. Adams writes as a practitioner reporting what she does (closing the laptop at ¶19), not as an authority prescribing. Short declaratives carry the strongest lines (¶18, ¶22). The register is warmest in the resignation section and most candid at ¶21; most management essays end on the pitch, this one ends on the limits. Hedges ("several times," "pretty strong signal," "pretty low industry bar") cluster where the evidence is thinnest, so the tone signals the weakness accurately. ## 🛡️ DARVO analysis Not applicable to this source. DARVO (Deny, Attack, Reverse Victim and Offender) describes a response to accusation, and this is a reflective essay with no accuser; the only challenge is the one Adams raises against herself at ¶21 and concedes. For completeness: **Deny** — nothing is denied; the essay volunteers its limits (¶21, ¶20) and declines to judge the other company at ¶2. **Attack** — no critics or former employers are attacked; the "grey" team appears only in the returner's words. **Reverse victim/offender** — none; the performance case at ¶12 is framed as something to handle "supportively," with no aggrieved manager. ## 🎯 Conclusion The durable idea is the design target. Retention is a goal managers and community leads only partly control; return is one you can design for, as Adams learned where pay was fixed and every hire left in eight weeks (¶7 to ¶8). Two tools follow and hold regardless of the evidence: the two-year test at ¶10 (what about this team would still be true after everyone has gone) as an audit question for any team or project, and the first-reaction rule at ¶18 (celebrate before you process) as a script anyone can write before they need it. The evidence is one returner's sentence, "several times" (¶4), and "the same result" (¶9) with no result named. Adams flags the survivorship problem herself at ¶21. Of the practices, one (¶14) rests on the anecdote alone, five rest on stated mechanisms you can test, and two (¶15, ¶20) appeal to unnamed evidence. Every link goes back to LeadDev. **Next steps for readers:** write the two-year test answer for your own team or program this week and count how many items are a person's name. Draft the first sentence you'll say the next time a contributor tells you they're leaving, and keep it apart from the handoff checklist. Before you cite any "people love coming back" story, ask who didn't come back and whether anyone asked them why. ## 🔄 Refactor Appendix — Build an engineering team people want to come back to > Density: 2–5 Literature atoms (H3 pad), 0–3 Permanent candidates (H4 pad). > Emitted: 4 Literature, 2 Permanent. Resonance filter applied — the Make memories > prescription (¶15) was dropped as uncited, and the stack-ranking practice (¶13) stays > in Notes as a practice rather than a claim. > > Konik claim-header rules: present tense, declarative, ≤10 words, searchable, unique. > > Atomicity: one declarative sentence per atom. "and"/"also" → split. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Split Pad #### [H1] A team is what survives its roster leaving > [!quote] [¶10] > Ask someone what a team is like and they’ll usually describe who’s on it. That’s a fair answer, but it has a short shelf life. **Context & use**: - ¶10, opening the first pillar; the four durable properties (standard, problems, teamwork, memories) follow in the same paragraph and become the next headings. - The two-year test is the instrument. Ask it of a DevRel program, a WordPress component, or a FAIR mirror: anything on the answer that's a person's name is a single point of failure. **Tags**: #xref #share #aphorisms <!-- atomicity: 1 idea ✓ --> #### [H2] Career-length feedback horizons shrink corrections > [!quote] [¶16] > It can help me say the hard thing earlier, when it’s still a small correction. **Context & use**: - ¶16, the second pillar; ¶17 adds the self-interested version (the manager will share a room with this person again in 5, 8, or 20 years). - Transfers to code review and maintainer mentoring: a reviewer who only comments on this PR misses the habit that blocks this contributor's path to maintainer. **Tags**: #xref <!-- atomicity: 1 idea ✓ --> #### [H3] The first reaction to a resignation is remembered > [!quote] [¶18] > The moment someone tells you they’re leaving is the moment they find out what you actually think of them. **Context & use**: - ¶18, opening the third pillar; the same paragraph says backfill and transition must happen but not first, and ¶19 supplies the script. - Scriptable for community moderators and team reps: write the first sentence before you need it, separate from the handoff checklist. **Tags**: #xref #share #articleseed <!-- atomicity: 1 idea ✓ --> #### [H4] A long relationship horizon is cheap to be wrong about > [!quote] [¶22] > Treating the horizon of our working relationship as long is cheap to be wrong about. **Context & use**: - ¶22, the closing argument; the asymmetry is spelled out in the next two sentences. - The decision-theory frame is what makes the essay transferable to volunteer communities, where "hire" never applied. Carry one correction with it: "nothing" is really "manager hours." **Tags**: #xref #aphorisms <!-- atomicity: 1 idea ✓ --> ### 🧠 Atomic Split Pad #### [P1] Design for return when retention is out of reach **Claim** (one sentence, Courtney's voice): When you can't control whether people stay, the variable you own is whether the stint was worth repeating, so design the program for the returner rather than the lifer. **Evidence pointer**: [[#H1]] · [[#H4]] · ¶7 to ¶8 (fixed pay, eight-week stints, returners cut training load) **Scope & conditions**: Holds wherever departure is structural: seasonal staff, open-source contributors, DevRel ambassadors, WordCamp organizers, fellowship alumni. Weakens for involuntary departures and for programs with no memory of who left, where the returner is re-onboarded from zero. **Links**: - [[Contributor lifecycle]] — because extends the onboard-and-retain model with a return stage - [[Retention as a community metric]] — because contrasts on what the program is optimized for - [[WordCamp organizer handbook]] — because applies-in organizer teams that turn over yearly <!-- atomicity: 1 idea ✓ --> #### [P2] Returner testimony only samples the people it worked for **Claim** (one sentence, Courtney's voice): Anyone who comes back is by definition someone the culture fit, so a story about returners says nothing about the people who stayed gone until you ask them too. **Evidence pointer**: ¶21 (the author's own caveat: a biased sample and the only sample that talks to her) **Scope & conditions**: Applies to every "people love coming back" testimonial, contributor spotlight, and alumni survey. Corrected only by collecting the non-returner side with the same instrument. ¶3's revealed-preference argument still holds for the returners; it doesn't generalize past them. **Links**: - [[Survivorship bias]] — because applies-in community health reporting - [[Every claim carries a fact]] — because extends to who the fact was collected from - [[Exit interviews for volunteers]] — because evidence-for the missing half of the sample <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Write the two-year test answer for the DevRel program and for The WPOCC fellowship design; count how many items are names — ¶10 #fleeting #inbox - [F2] Draft the first sentence for a contributor step-back announcement in community channels, separate from the handoff checklist — ¶18 to ¶19 #fleeting #inbox - [F3] Open question: does CHAOSS or any WordPress contributor dashboard track returners (inactive 6+ months, then active) as a class? If not, propose the metric — ¶7 #fleeting #inbox - [F4] Find one primary source for the novelty-and-memory claim ¶15 gestures at before repeating the "make memories" advice anywhere #fleeting #inbox - [F5] Article seed: "Retention is for employers; return is for communities," with the ¶8 retail experiment as the framing story — https://leaddev.com/culture/build-an-engineering-team-people-want-to-come-back-to #fleeting #inbox #articleseed - [F6] Check Daniel Terhorst-North's own fluid-teams writing (dannorth.net) against what ¶13 attributes to it before citing him through this article #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Write the two-year test for the DevRel program and list which durable properties are role-bound versus person-bound — program design, 30 min #things - [ ] Draft a scripted first response to contributor step-back announcements and share it with community moderators — community ops, 20 min #things <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Build an Engineering Team People Want to Come Back To.md|Build an Engineering Team People Want to Come Back To by Meg Adams]] ## Highlights - Ask someone what a team is like and they’ll usually describe who’s on it. That’s a fair answer, but it has a short shelf life. What I’m after is the part of the description that would still be accurate two years after everyone they worked with has gone – the standard of performance, the challenging and exciting problems, the teamwork, and the memories created. ([View Highlight](https://read.readwise.io/read/01m3a1qytz0wx3szz8cfpv01cm)) - Note: [H1] A team is what survives its roster leaving - Tags: atom/literature/h1 - Performance management loops are designed largely for managers to speak to the current cycle, which is useful in terms of specificity. However, when we stretch the horizon, often the feedback we want to deliver changes shape. It can help me say the hard thing earlier, when it’s still a small correction. I get specific about what might be working well for someone now, but I can already see that it will hold them back two levels from now. ([View Highlight](https://read.readwise.io/read/01m3a1r0z744gmexsfmtv8mc7p)) - Note: [H2] Career-length feedback horizons shrink corrections - Tags: atom/literature/h2 - The moment someone tells you they’re leaving is the moment they find out what you actually think of them. Most of the machinery around resignations treats them as a gap to close, with processes to kick off for backfill, transition plans, and coverage. All of that indeed has to happen, but none of it should happen first, because the first thing is the only part anyone remembers. ([View Highlight](https://read.readwise.io/read/01m3a1r3dfwnzd5cdhqk436j1b)) - Note: [H3] The first reaction to a resignation is remembered - Tags: atom/literature/h3 - Matching a compensation number isn’t always in your control, but designing teams people want to return to can be. Treating the horizon of our working relationship as long is cheap to be wrong about. If I treat someone as a colleague for the next 20 years and they never speak to me again, I’ve lost nothing. If I treat them as a two-year hire and I’m wrong, I’ve spent those years skipping the conversations that would have made them, and us, better. ([View Highlight](https://read.readwise.io/read/01m3a1r5q617mjxknd2bxdnfjz)) - Note: [H4] A long relationship horizon is cheap to be wrong about - Tags: atom/literature/h3