How to Run a Workplace Mental Health Assessment: A Step-by-Step Guide

Most organizations that say they want to "assess mental health" are asking two different questions at once. One is about people: how are our employees doing? The other is about the organization: what is it about how we work that helps or harms them? The first question belongs mostly to clinicians and to employees themselves. The second is the employer's, and it is the one a workplace mental health assessment should answer.

That distinction shapes everything that follows. An organizational assessment looks at conditions, policies, leadership behavior and access to support. It does not diagnose anyone, rank individuals or produce a list of people who seem to be struggling. If a project starts drifting toward any of those outcomes, stop and reset. Individual screening raises legal and ethical questions under the ADA and related rules, and it is a separate subject.

The process below suits a mid-sized employer running its first assessment. It is written in stages because the order matters more than the tools. Teams that jump straight to a survey usually end up with data nobody trusts and no plan for what to do with it.

The stages at a glance

Stage	Main question	Typical owner	Rough duration
1. Scope	What decisions will this inform?	HR lead plus executive sponsor	1–2 weeks
2. Inventory	What do we already know?	HR, benefits, people analytics	1–2 weeks
3. Design	What will we ask, and of whom?	People analytics with a cross-functional group	2–3 weeks
4. Protect	How do we keep responses safe?	Privacy or legal, survey vendor	Alongside design
5. Field	How do we get honest participation?	Internal comms, managers	2–3 weeks
6. Analyze	Which patterns matter?	People analytics	About 2 weeks
7. Report and act	What will we change, and when?	Executive sponsor, HR	4–6 weeks to first actions
8. Re-measure	Did anything move?	People analytics	9–12 months later

The durations assume an organization of a few hundred to a few thousand people. A 60-person firm can compress most of this into a month. A large multi-site employer will need longer for stages 3 and 5.

Stage 1: Scope the assessment around decisions

Start with decisions, not questions. Sit down with the executive sponsor and write out, in plain language, what the organization would be willing to change depending on the results. More EAP sessions? Manager training? Changes to scheduling or on-call rotations? A revised leave policy?

If the honest answer is "nothing much," the assessment is premature. Asking people about their working conditions and then doing nothing is worse than not asking, because it teaches employees that speaking up is pointless. That lesson carries into every future survey.

A useful scoping document fits on one page and covers:

• the purpose, in a sentence employees could read without wincing
• the population (everyone, or specific sites or functions)
• the areas to be covered
• who will see what level of data
• the date by which results will be shared back with employees

A simple framework keeps scope honest. Our four key areas (awareness and stigma, culture and climate, access to care and support, and leadership commitment) map well onto things an employer can actually change. The US Surgeon General's Framework for Workplace Mental Health and Well-Being is another solid reference, organized around five essentials such as protection from harm and mattering at work.

Stage 2: Inventory what you already have

Before writing a single survey item, gather the data the organization already holds. Much of the picture is usually sitting in different systems.

Typical sources include aggregate EAP utilization reports, short-term disability claims by category (aggregate only, from the carrier), voluntary turnover and exit interview themes, absence patterns, overtime and schedule data, engagement survey results, and trends in grievances or HR cases. None of these measure mental health directly. Together they can show where pressure is building, and they give you a baseline that does not depend on anyone filling in a form.

Keep everything at the aggregate level. You are looking for patterns across teams and time periods, not individual records.

This is also a good point for a small group of leaders to complete our workplace mental health self-check. It takes about five minutes and scores the same four areas. Leaders' answers are no substitute for employee data, but comparing the two later is often revealing. When leadership rates the culture far more favorably than staff do, that gap is itself a finding.

Stage 3: Design the instrument

Now decide what to ask. A few principles hold up well across organizations.

Ask about conditions, not symptoms. Items like "I have enough time to do my work properly" or "My manager would respond well if I said I was struggling" describe the workplace. Clinical screeners such as the PHQ-9 or GAD-7 describe individuals' health and belong in a clinical setting. Leave them out of an employer-run assessment.

Keep it short. Twenty to forty items is a reasonable range. Every additional item costs completion, and long surveys produce tired, patterned answers toward the end.

Borrow established items. Validated instruments for psychosocial working conditions exist, including the UK Health and Safety Executive's Management Standards Indicator Tool and the Copenhagen Psychosocial Questionnaire. Well-tested wording, used within its licensing terms, beats anything invented in a conference room.

Limit open-text questions. Two or three is enough. "What is one thing that would make your work less stressful?" yields more usable material than a page of free-text boxes.

Test the draft. Run it past a small group from different roles, including frontline and shift staff if you have them. Watch for jargon, double-barreled items and anything that reads as though it might identify someone.

Stage 4: Build in confidentiality

Confidentiality is not a footnote to the design. It is the condition under which people tell you the truth.

Set a minimum reporting group size and hold to it. A common rule is to suppress any breakdown with fewer than five respondents, and many organizations set the bar at ten. Be careful with demographic questions: tenure, site, shift and role combined can point to a single person in a small team even though no name was collected. Ask only for the attributes you will genuinely use.

Use a third-party survey platform, or at minimum make sure line managers cannot see raw responses. State in the invitation exactly who will see what. If you promise anonymity rather than confidentiality, make sure it is literally true, with no login tracking and no unique links tied to employee IDs.

Under the ADA, medical information an employer obtains must be kept confidential and separate from personnel files. A well-designed organizational assessment should not collect medical information at all, which is one more reason to avoid symptom questions.

Stage 5: Field the survey

Participation depends on trust, and trust depends heavily on what happened after the last survey. If there is a history of surveys that went nowhere, acknowledge it in the launch message.

Practical steps that help:

1. Have the executive sponsor send the invitation in their own words, saying what will happen with the results and by when.
2. Brief managers beforehand so they can answer questions without pressuring anyone.
3. Give frontline and shift staff paid time and a device to complete it. Assuming everyone reads email at a desk is the most common reason participation skews toward office staff.
4. Keep the window to two or three weeks, with one or two reminders.
5. Share participation rates by area during fielding, never individual completion.

End the survey with support information: the EAP, the health plan's behavioral health line, and the 988 Suicide & Crisis Lifeline, which anyone in the US can call or text at 988. A survey about wellbeing will sometimes reach someone at a hard moment.

Stage 6: Analyze for patterns that matter

Resist the urge to produce a single score. An overall "mental health index" is easy to report and nearly useless for deciding what to do.

Look instead at:

• Area-level results. Which domains are weakest overall?
• Variation between groups. Differences between sites, shifts or functions are usually more actionable than the company average.
• Gaps between perception and experience. Compare leaders' self-check answers with employee results, and compare survey data with the operational data from stage 2.
• Open-text themes. Code comments into categories and count them. A theme raised by many people in their own words carries weight.

Be honest about uncertainty. Small groups produce noisy results, and a three-point difference between two teams of twelve probably means nothing. Flag low-confidence comparisons rather than presenting every number as precise.

Stage 7: Report back and commit to actions

Share results with employees within the timeframe you promised, ideally within six weeks of the survey closing. Include the uncomfortable findings. People already know what is wrong; leaving it out of the summary only signals that leadership is not ready to hear it.

Pick a small number of actions, usually two to four, and give each an owner and a date. An illustration helps show the scale.

Take a hypothetical 300-person distribution company with two warehouses. Its results show reasonable scores on access to care but low scores on workload and on managers' confidence discussing mental health, concentrated on the night shift at one site. Its actions might be: revise night-shift staffing for peak season (operations director, by next quarter); train all shift supervisors to recognize distress and hold a first conversation (HR, within four months); and move the EAP briefing into night-shift huddles instead of daytime emails (benefits lead, next month). That is a manageable list, tied directly to the data, with obvious ways to tell whether it happened.

The WHO's guidelines on mental health at work are a useful check at this stage. They place organizational interventions, which address working conditions directly, alongside manager training and individual support. If your action list is mostly resilience workshops, it is probably leaning too hard on the individual. For more ideas by area, see the recommended actions under leadership commitment and the tools in our resource library.

Stage 8: Re-measure and close the loop

Plan the second assessment before the first one ends. Nine to twelve months is a typical interval: long enough for changes to take hold, short enough that people remember the first round.

Keep the core items identical so the results can be compared. Add a short section asking whether employees noticed the actions you committed to. If they did not, that tells you something about communication, or about whether the actions were real.

Over two or three rounds, the assessment stops being a one-off project and becomes part of how the organization runs.

Mistakes that undo good work

• Treating the assessment as a wellness campaign rather than a diagnostic of working conditions.
• Collecting symptom or diagnosis data an employer has no business holding.
• Breaking results down so finely that individuals become identifiable.
• Reporting averages and ignoring variation between teams.
• Announcing actions that nobody owns.
• Letting a year pass with no visible change, then launching another survey.

None of these takes specialist expertise to avoid. It takes deciding, at the start, that the assessment exists to change something, and then building every stage around that decision.