Career Growth
Make Glue Work Visible in Your Engineering Performance Review
Trace a cross-team technical contribution without claiming collaborators' work, then use two completed cases to align recognition and future glue-work allocation with actual role expectations.
Make glue work visible by tracing your specific intervention to a verifiable change in another person's work, while preserving their ownership of the result. Identify the technical judgment, coordination artifact, or follow-through you supplied; show what the team did with it; and ask your manager how that contribution fits your actual role expectations. Being helpful is too vague to evaluate, and taking credit for the whole project is inaccurate.
This guide from LandOffer.ai uses original fictional engineering cases. The people, documents, conversations, and role agreements are teaching examples, not employee records or measured workplace outcomes. The focus is attribution and future work allocation within a performance review, rather than a general annual self-evaluation or a promotion packet.

Identify the contribution hidden inside someone else's result
Tanya Reilly's original Being Glue describes work such as unblocking colleagues, reviewing designs, onboarding, and improving processes. She explains how deliberately managed glue work can support technical leadership, while unexamined expectations can create career problems. That practitioner account gives useful language for a pattern; it does not establish how every employer evaluates collaboration.
For a review, replace “I kept everyone aligned” with the particular uncertainty you resolved. Did two teams assume different meanings for a field? Was a release dependency missing an owner? Did a colleague need a documented decision rather than another meeting? Describe what you contributed at that point and which artifact lets a reviewer inspect it.
Separate technical judgment from attendance. Scheduling a meeting may have been necessary, but the more useful evidence might be the incompatible assumptions you brought to that meeting and the decision record you produced afterward. Conversely, do not relabel routine logistics as architecture simply because an engineering project benefited from it.
Write a contribution claim that another participant can verify and your manager can compare with the role expectations. It does not require a dramatic savings number or a claim that everything would have failed without you.
Completed asset: trace one intervention through shared work
In our fictional case, senior engineer Inez helps a catalog team and a web team agree on their integration-test fixture interface. The catalog fixture specification uses product_id; a web smoke-test example expects sku_id. Inez identifies the mismatch during a design review and assembles both references in one comparison note.
The teams agree that the web fixture loader should map the producer's field explicitly. Milo owns implementing that change. Bea, the QA lead, owns defining the relevant check coverage. Inez records the decision and handoff. This scenario stops at an agreed interface and assigned implementation work; it does not claim a deployed patch or a passing test suite.
Here is the completed contribution map:
| Evidence artifact | Inez's contribution | Another person's responsibility or result | Claim boundary |
|---|---|---|---|
| Interface comparison | Places the catalog product_id definition beside the web sku_id expectation and names the mismatch |
Catalog and web leads confirm the intended producer-consumer contract | Identifies a concrete inconsistency; no measured failure reduction |
| Decision note | Records the agreed web-loader mapping and the producer contract left unchanged | Both leads own their teams' agreement; Milo owns implementation | Inez records and follows through on the decision, not sole design ownership |
| Handoff record | Names Milo for the loader change and Bea for coverage definition, with an open question about the empty-catalog case | Milo and Bea retain the respective engineering and QA work | Assigned work is not completed work |
| Review follow-up | Returns the unresolved empty-catalog question to the two owners and updates its status to pending | The owners must decide expected behavior before closing their work | Follow-up is visible; resolution and test completion are not invented |
The four artifacts let a reviewer trace the intervention without assigning every later result to Inez. The comparison establishes what Inez noticed. The decision note shows how the group responded. The handoff preserves the people doing the next work. The follow-up prevents an unresolved edge case from disappearing behind a meeting summary.
This is stronger than collecting every calendar invitation. It also protects collaborators from having their implementation work absorbed into a broad claim that Inez “delivered the integration.” Her work belongs in the review without rewriting who did what.
Write the review passage with attribution intact
A finished review passage for Inez could read:
I identified a mismatch between the catalog fixture's
product_iddefinition and the web smoke-test example'ssku_idexpectation. I wrote the comparison used in the joint review, recorded the leads' agreement to adapt the web loader, and documented the handoff to Milo for implementation and Bea for coverage definition. I also kept the empty-catalog behavior question visible as pending rather than marking the interface work fully closed.The shared result was an agreed fixture interface and named owners for the remaining engineering work. My contribution was the cross-team analysis, decision record, and follow-through. Milo owns the loader implementation; Bea owns the coverage definition. We have not measured test-failure reduction, delivery-time savings, or production impact from this work, and this record does not establish a shipped change.
The passage gives a reviewer a useful reason to inspect the underlying material. It names the judgment and resulting decision, credits others, and states the limit. It does not depend on a praise quote saying Inez was indispensable.
If the later implementation is completed, update the evidence with its actual status and attribution. Do not silently upgrade the original coordination claim into authorship of the patch. The relationship between your work and the later result should remain visible even when both belong in the same project narrative.
Completed case one: keep useful glue work and define its recognition
Inez's fictional team expects senior engineers to clarify cross-team technical dependencies and document ownership. Her manager, Arun, agrees that this fixture work fits that current-role expectation. This is a stated assumption of the case, not a universal senior-engineer requirement.
Published frameworks illustrate why the role check matters. GitLab's senior infrastructure framework includes mentoring, improving processes and runbooks, and providing information to prioritize work. These are expectations for that named company and function. They do not make every coordination task senior-level evidence everywhere.
In the completed fictional discussion, Inez asks: “Which part of this work should appear in my review, and what implementation work do you still expect me to own?” Arun's recorded response is specific: “Use the interface comparison and handoff as evidence for our cross-team technical follow-through expectation. The loader implementation belongs to Milo. This example does not replace your own planned service changes or establish performance at another level.”
Their resulting allocation agreement is also filled in. Inez will finish the existing handoff record and one scheduled follow-up on the pending edge case. Milo and Bea keep their named responsibilities. Arun will route new coordination requests through sprint planning rather than assuming Inez accepts them by default. Inez's next service change stays on her work plan. They will review the allocation on October 30, 2026; that is a future check, not a completed result.
The review record now identifies what is recognized and what remains expected. It promises no rating, promotion, or compensation outcome. If Arun cannot identify an applicable role expectation, that uncertainty belongs in the discussion rather than being concealed by an impressive-sounding label.
Completed case two: acknowledge the contribution and redistribute the load
A second, independent fictional team has a different problem. Early-career engineer Eli receives five informal onboarding requests: two local-setup questions, two questions about fixture ownership, and one access request. The fictional request record shows Eli supplied existing setup references, wrote a fixture-owner index, and routed the access request to the authorized owner. Eli does not grant access.
Those are useful contributions. They do not establish that five colleagues became productive faster, that all five requests were resolved, or that Eli prevented a security incident. The evidence is the request record, the references supplied, the owner index, and the access handoff. The broader effect has not been measured.
Eli's manager, Jo, also observes that the team's stated expectation for Eli includes owning a bounded parser change through implementation and review. The support requests have displaced the planned development sessions. Their review needs to recognize the assistance and repair the allocation, rather than praising Eli for absorbing an unlimited support role.
Here is the completed fictional agreement:
Review recognition: Record Eli's setup references, fixture-owner index, and appropriate access handoff as collaboration evidence. Credit the maintainers who wrote the setup instructions and the resource owner who decides access. Do not describe Eli as author of those instructions or owner of access approval.
New allocation: Jo routes onboarding requests to a visible rotation. Sam takes the next week's intake and Mei the following week's. Eli supplies the owner index, then participates in the ordinary rotation after the first parser-change review. No one becomes the permanent default because they answered the last question.
Development expectation: Jo protects two scheduled implementation sessions for Eli's parser error-case change and names Sam as its reviewer. Completion still requires the normal implementation and review evidence; the sessions themselves are not proof of delivery.
Follow-up: Jo and Eli will compare the actual request distribution and parser-review status at their next scheduled check-in. They have not yet observed a reduced support load or completed parser change under this agreement.
Eli's review record needs the allocation agreement alongside the contribution evidence. Jo names Sam and Mei for the next intake shifts and preserves Eli's implementation sessions, making the allocation inspectable. Recognition and redistribution can both be necessary.

Ask collaborators to verify the facts, not endorse your rating
A corroboration request should be narrow enough to answer honestly. For Inez, a suitable message is: “I plan to say I identified the field-name mismatch, wrote the comparison, and recorded the handoff. Does that match your recollection? Please flag anything I have attributed incorrectly. I will credit Milo's implementation and Bea's coverage work separately.” This is sample text, not a message we sent.
Ask before quoting private feedback, and use the review channels your organization allows. A manager may have a process for gathering partner input. Do not turn a request for factual confirmation into pressure to say you deserve a particular rating or level.
Dropbox's engineering career-framework overview says its framework is not a promotion checklist or an exhaustive list of behaviors, and calls for alignment with the manager on role expectations and impact goals. That company-specific framing supports asking what your employer actually evaluates, rather than importing another organization's wording as an entitlement.
If a collaborator disagrees, revise the attribution or preserve the disagreement for your manager to assess. Use the confirmed contribution and identify any disputed attribution for your manager to review. Keep the original artifact available so the conversation does not rely only on memory.
Keep unmeasured impact and future responsibility separate
Use observed intermediate results when they are the evidence you have. A dependency gained an owner; a design assumption was made explicit; a handoff documented the remaining question. Those are meaningful facts within a project. Do not convert them into invented revenue protected, outages avoided, or hours saved.
When you have reliable measurements, explain the source and what they measure. Meeting counts measure meetings, not leadership quality. Fewer support messages might reflect a routing change rather than better documentation. Distinguish the observation from your hypothesis about why it occurred.
Reilly's original account also emphasizes artifacts and deliberate career conversations. Use that idea to surface an actual allocation decision: what work should you continue, what should rotate, what technical responsibility must remain yours, and what evidence will your manager consider? Being good at coordination does not require choosing a different career path by default.
Bring one contribution chain and one proposed allocation change to the review. Leave with the agreed attribution, applicable role expectation, next owners, and an explicit point to revisit unfinished work. The review becomes useful when it describes both the work you enabled and the work you are expected to own next. For further workplace reading, visit LandOffer.ai.
Sources and Further Reading
- Tanya Reilly: Being Glue — original practitioner discussion of glue work, artifacts, role expectations and career choices.
- GitLab: Senior Infrastructure Career Framework — a specific employer and function's published expectations.
- Dropbox Engineering Career Framework — manager alignment and limits of treating a framework as a checklist.


