People Management
Run a Team Health Check and Agree on One Improvement to Test
A fictional seven-person engineering team discusses four ongoing working conditions and selects a small knowledge-transfer trial. Completed voting, capacity, item-review, and follow-up records show how to retain disagreement, distinguish completed effort from future commitments, and choose a bounded recheck without claiming the team is fixed.
Run a team health check by asking people about recurring working conditions, preserving different experiences, and choosing one improvement the team can actually try. Record who will do what, what capacity they have accepted, and when you will examine both the experience and the evidence. A change in vote colors alone does not establish that the working condition improved.
This LandOffer.ai guide works through a fictional engineering team’s discussion and completed follow-up. The aim is a useful conversation about how the team operates across projects. It is not an individual performance assessment, a medical diagnosis, or an anonymous employee survey.

Set a clear purpose before asking for ratings
A project retrospective usually looks back at a bounded piece of work or recent period. A health check can examine conditions that persist as projects change: whether people can obtain help, understand priorities, learn the systems they support, or raise concerns. Choose a small set of relevant prompts rather than importing every category from someone else’s framework.
Spotify’s March 2023 practitioner account describes this broader perspective and advises tailoring questions, exploring experiences, and following through. It also warns that rushing to actions can crowd out understanding. Its experience is useful guidance, not proof that copying its workshop will improve your team.
Before inviting people, write the purpose, participants, time window, and record audience. Explain whether managers will attend, whether individual votes will be visible, and what leadership will receive. Do not promise anonymity in a seven-person conversation where colleagues can recognize an incident. Agree on what can be recorded without names and what needs a separate private route.
If a participant raises harassment, a serious safety issue, or another concern needing an established reporting process, do not turn it into an experiment to be voted on. Explain the appropriate route and recording limits. A facilitator cannot guarantee confidentiality or resolve every employment concern through a workshop.
Worked health check: keep the disagreement in the record
The following people, workplace, events, and results are fictional. Cedarline’s seven-person internal-tools team maintains several services. Over six weeks, responsibilities have moved between pairs of engineers, but specialist knowledge remains concentrated. Manager Lena asks facilitator Jules, who has no reporting relationship to participants, to run a 60-minute discussion.
The invitation states: “We are checking whether our everyday working arrangements support the team across services. We will retain combined vote counts and agreed examples in our team workspace. Lena will receive the same summary. Individual voting notes will not be retained. Speaking here is not anonymous; you may pass or request a separate conversation.” All seven attend and choose to vote on each prompt.
Jules reserves five minutes for purpose and boundaries, eight for quiet reflection and voting, 27 for discussing differences, 15 for choosing and shaping a possible trial, and five for closing. That is a proposed local agenda, not a provider requirement. If a deeper concern needs more time, the group can schedule a continuation rather than manufacture agreement.
For this fictional session, green means generally working in the stated period; yellow means uneven or needing attention; red means repeatedly obstructing work. These descriptions support conversation, not numerical scoring. Participants reveal their choices together, then describe examples without naming a colleague as the problem.
| Prompt about the previous six weeks | Green / yellow / red votes | Experiences retained in the summary | Interpretation still open |
|---|---|---|---|
| We can identify the current service contact and get a useful response | 4 / 2 / 1 | Established service contacts are clear; contact lists lag after responsibility changes | Is the main gap the list, availability, or unclear authority? |
| We can learn enough to take over a routine support question | 1 / 3 / 3 | Written explanations exist, but the reasoning behind common decisions is hard to find | Documentation volume does not establish usable knowledge transfer |
| We understand what takes priority when requests compete | 3 / 3 / 1 | Weekly priorities are visible; urgent requests sometimes arrive without an explicit trade-off | The team cannot change another department’s escalation rules alone |
| We can raise an uncertain answer without being blamed | 5 / 1 / 1 | Most describe helpful responses; one person says a public correction felt dismissive | Majority comfort does not invalidate that experience |
Atlassian’s Health Monitor uses different colored votes followed by discussion and improvement planning. It explicitly cautions against comparing teams. This example adapts the method; it does not reproduce an official assessment or assign Cedarline a validated health score.
The learning prompt draws the most red votes, but Jules does not declare it the winner automatically. Priya explains that knowing where a document lives is different from understanding when an exception applies. Ben, who voted green on contact access, agrees that finding him is easy while getting a self-contained explanation often requires a call. Another participant says a lengthy new documentation requirement would make that worse by consuming support time.
Jules records: “Knowledge transfer is uneven even where contact access works. The team wants to test a small explanation format before proposing a documentation overhaul.” The red vote on speaking up remains in the summary. With the participant’s agreement, Lena offers a separate conversation; the summary records an offer, not acceptance or resolution. The group does not bargain away that concern to obtain one action.
Turn one concern into a bounded trial
Selecting one experiment limits the work, not the importance of other concerns. Ask whether the proposed change addresses the experience people described, whether the team controls it, and what existing work will make room. “Improve collaboration” is too broad to assign. “Write a complete manual for every service” exceeds the problem and available capacity.
Cedarline selects a short explanation attached to a routine support answer. It will include the question, the reasoning for the chosen answer, and the condition that would require escalation. This is meant to help another engineer understand a decision, rather than merely copy the response.
The team declines a new recurring training meeting for now because it would require attendance from specialists already covering support. It also postpones a service-wide ownership redesign: Lena can update the team’s contact information, but changing other teams’ responsibilities requires their agreement. Neither item is presented as a commitment made elsewhere.
The trial record is completed before work begins:
| Field | Cedarline’s agreed trial |
|---|---|
| Working hypothesis | A short reasoning note may help a second engineer explain a routine answer without immediately asking its author |
| Scope | Four routine support questions selected during November 2–13, 2026; exclude incidents, sensitive access requests, and policy exceptions |
| Owner and accepted capacity | Ben agrees to create four notes, up to 10 minutes each; Priya agrees to review them, up to five minutes each |
| Work displaced | Lena removes one 60-minute internal documentation tidy-up from that period; urgent support retains priority |
| Review method | Priya attempts to explain each answer and its escalation condition using the note; Ben records missing information, not a score for Priya |
| Burden and safety check | Stop adding a note when its author reaches 10 minutes; do not include customer details, credentials, or restricted incident material |
| Recheck | November 16: examine all four records, ask about usefulness and burden, and revisit the same learning prompt |
| Decision choices | Continue unchanged, revise a named weakness, stop, or gather more information; no automatic rollout |
The capacity limit is 40 minutes for Ben plus 20 for Priya, matching the 60 minutes Lena displaced. It is an allocation, not a claim of time saved. The trial checks comprehension of four explanations; it cannot establish readiness to operate a service independently or handle an emergency.

Recheck what happened, including what did not work
Bring the original concern and trial boundaries to the follow-up so that completing notes does not become proof that knowledge transfer improved. Otherwise a completed activity can quietly become a claim of improvement. The GOV.UK Service Manual advises teams to agree changes and use later discussions to check whether they addressed the problems raised. That is general delivery guidance, not a prescribed health-check cadence for every employer.
In the fictional November 16 review, Ben and Priya bring the following record. Each assessment concerns a note’s explanation, not production execution or an employee’s general competence.
| Trial item | What the review found | Disposition |
|---|---|---|
| R1: selecting an existing report period | Priya explained the selection and the boundary requiring another owner | Keep the concise explanation as an example |
| R2: identifying a duplicate request | The note described the decision but omitted where to find the request identifier; Priya asked Ben | Add the identifier location; revised note has not been rechecked |
| R3: choosing a routine export format | Priya explained both the choice and the exception requiring escalation | Keep; no claim that an export was executed |
| R4: finding the approved support contact | Ben reached the authoring limit while checking conflicting contact references | Stop the note; Lena must establish the authoritative reference first |
Ben used 10, 10, eight, and 10 minutes: 38 minutes. Priya used five minutes on each record: 20 minutes. Total recorded effort was 58 minutes, within the allocation. R4 still consumed review time because Priya examined the incomplete note and recorded the unresolved reference. It is not counted as a completed explanation.
The repeat learning prompt receives two green, four yellow, and one red vote from the same seven participants. The group does not describe this as a measured improvement rate. Different work occurred during the trial, the questions were selected locally, and people may interpret the colors differently. One participant remains concerned that notes will become another obligation on overloaded specialists.
The completed decision reads: “Revise and run one further bounded check; do not roll out. Priya accepts five minutes to recheck revised R2 during the next weekly support slot. Lena accepts 15 minutes to reconcile R4’s contact reference before any new note is requested. Ben accepts no additional authoring commitment until that reference is settled. Revisit on November 23.” Those 20 future minutes are newly agreed work, not part of the completed 58-minute total.
The concern about transferring decision reasoning across services is still open. Two notes explained their cases adequately in this limited review, one needs a recheck, and one exposed a prerequisite. These findings support a limited recheck, not a conclusion that knowledge sharing is fixed.
Share a useful summary without turning it into a ranking
Send the agreed record to its stated audience: conditions discussed, meaningful disagreements, selected trial, accepted capacity, follow-up evidence, and unresolved dependencies. Give participants a chance to correct factual errors before broader sharing; a silent response is not permission to share a personal account. Keep personal accounts out of a leadership slide unless the relevant sharing has been agreed and is appropriate under your organization’s processes.
If leadership needs an organizational view, clarify that purpose in advance and share the minimum useful information. Do not publish a league table of red votes, interpret a small group’s chart as an engagement survey, or reward managers for converting yellow to green. That would change the incentives for honest discussion.
For your next health check, choose a few ongoing working conditions and ask what experiences explain the differences. Leave with one feasible trial only when the team understands and accepts it; if more exploration is needed, record that next step instead. Keep the original concern beside the follow-up evidence. More workplace guides are available from LandOffer.ai.
Sources and Further Reading
- Atlassian: Team Health Monitor, voting, discussion, action ownership, and follow-up — checked October 27, 2026; practitioner framework, not a validated individual assessment.
- Spotify Engineering: Getting More from Your Team Health Checks — March 6, 2023; tailoring, systemic discussion, and follow-through based on Spotify’s experience.
- GOV.UK: Core principles of agile — current public body checked October 27, 2026; page last updated May 23, 2016; agree changes and revisit prior problems.


