Short answer
Qualitative engagement data consists of employees' descriptions, examples, and interpretations of their work experience. It may come from survey comments, interviews, group discussions, or other agreed listening channels. Analysis turns that material into themes whose supporting evidence and limits are visible. A theme can suggest a useful question or action; it does not establish the cause of a score or predict an individual's departure.
Define the question and the material
Start with a question narrow enough to answer. “What makes handovers difficult after the scheduling change?” is more useful for a focused review than “Why are employees disengaged?” Specify the population, period, channels, and intended decision. Keep excluded material and missing groups visible.
Open survey comments are already qualitative data. A conversation can provide more detail when appropriate, but it is not automatically more truthful. Different channels can attract different participants and produce different accounts. Retain that context when interpreting the material.
The UK government's quality framework for qualitative evaluation asks reviewers to examine how conclusions follow from evidence, how participants were included, and what limits apply to wider inference. The practical worksheet below is an editorial example for a workplace review, not a claim that a short HR exercise is a formal research study.
Build a small coding guide
A code is a label applied to a relevant part of an account. A theme is an interpretation that connects codes around the question being investigated. Do not treat a common word such as “manager” as an explanation.
For a first pass, define each code, its inclusion rule, and a nearby exclusion. Add new codes when the material does not fit, and note why the guide changed.
| Proposed code | Include | Keep separate |
|---|---|---|
| Priority conflict | Two instructions that cannot both be followed | A high workload with clear priorities |
| Handover access | Difficulty finding the current handover information | Disagreement with a clearly available instruction |
| Escalation uncertainty | Not knowing who can resolve an exception | Knowing the owner but finding them unavailable |
| Helpful practice | A concrete routine the participant says helps | General praise with no work example |
These codes are illustrative. They are not a universal engagement taxonomy or a product classification.
Work through a fictional example
The following excerpts are invented for this guide.
| Fictional excerpt | Initial code | Question to check |
|---|---|---|
| “The morning list says returns first, but the shift message says replenishment.” | Priority conflict | Were the messages issued for the same shift? |
| “The latest list is clear when I can open it, but the shared tablet is often elsewhere.” | Handover access | Which shifts and access arrangements are affected? |
| “Our lead spends two minutes confirming the list before opening.” | Helpful practice | Under what conditions is that routine feasible? |
A draft theme could be: teams describe inconsistent access to the current priority list. That interpretation needs checking against the original material. The first excerpt may reflect conflicting instructions rather than an access problem. Preserve this alternative instead of forcing all three excerpts into a single story.
A reviewer might separate the theme into instruction consistency and access to the agreed version. That produces two possible owners and avoids recommending a communication workshop for what might partly be an equipment issue.
Review the analysis before reporting it
- Read the relevant context around each excerpt, not only the extracted sentence.
- Look for accounts that contradict the proposed theme or describe an effective practice.
- Check whether one prolific participant dominates the material.
- Record which groups or shifts are absent and what cannot be inferred about them.
- Where reviewers disagree, document the difference and resolve it against the material or retain the uncertainty.
- Remove unnecessary identifying detail from the report; keep access to raw material restricted to authorised reviewers.
Frequency can describe the material reviewed, but comments are not automatically independent people. If one participant mentions an issue five times, do not count that as five employees. Unless the sampling and denominator justify it, “mentioned in these accounts” is more accurate than “experienced by 40% of staff.” An unusual but serious concern may also require a separate response even when it is not frequent.
Turn a theme into an action brief
Use a short record the action owner can work with:
| Field | Fictional example |
|---|---|
| Observation | Some reviewed accounts describe difficulty finding the current priority list |
| Coverage and limits | Feedback from participating shifts only; other shifts not represented |
| Alternative explanation | Some instructions may conflict even when accessible |
| Check before action | Compare current instructions and ask how each shift accesses them |
| Proposed test | Make one current list accessible at the agreed handover point |
| Owner and review | Store operations owner; review after the next scheduling cycle |
| Feedback to participants | Explain the test, its limits, and how to report whether it helps |
Compare later feedback using the same question and a stated population. A fall in mentions may reflect improvement, fewer responses, or a changed collection method. It is not proof of success on its own.
Use AI as an aid to a reviewable process
If an AI tool suggests codes or summaries, test those suggestions against the source material. Check missed exceptions, invented explanations, and whether different phrasing changes the result. Keep sensitive employee material within the tools and permissions approved for that purpose. An AI-generated label is a draft interpretation, not a verified finding.
Lontra can support preparation of employee conversations and review of relevant work context. Start by exploring the fictional team diagnostic if the need is to investigate one concrete team question. For the broader decision process, read people analytics beyond dashboards; for timing and response ownership, see real-time employee engagement.