How we size productivity loss and recovery

Chapter 1 The calculator joins two studies and two values from outside them

The loss comes from a study of working Poles (2026), the recovery from a study of Mindgram users (2025). The calculator joins the two numbers in one calculation.

Chapter 2 Reduced mental health costs the equivalent of 3.6 working days a month

The 2026 Polish Workers' Productivity Study was fielded by Experience Institute, commissioned by Mindgram. ^{Source 1} It covered 2,302 working adults from seven sectors.

Presenteeism is more than four times larger than absence: people are at work, but working worse.

2.1 Three answers become an equivalent in days

WPAI asks about the last seven days. From the three answers we compute, for each person, the monthly equivalent of days of reduced productivity:

2.2 We count only the excess above people with no symptoms

Fully healthy people declare some loss of output too. So from the average we subtract the level of people whose all three DASS-21 subscales are within the norm.

Mindgram's own computation on the 2026 Polish Workers' Productivity Study data, weighted by age and sex; n = 2,159 after dropping 143 records with implausible hours or non-numeric entries. This number is not in the report.

Under other assumptions the result sits between 3.0 and 4.3 days. Adjusting for age, sex, sector and work model gives 3.4 days.

Chapter 3 Active users cut the loss by 0.8 of a day a month more than inactive ones

This is a user study with a comparison group. ^{Source 2} It covered 272 employees of client companies. In two waves, between March and September 2025, they completed DASS-21 and WPAI.

Active users drew on individual support on the platform between the waves: sessions with a specialist, expert consultations or AI-assisted programmes. Inactive users used no form of support. We count the change among the active minus the change among the inactive.

Two numbers we use on the site follow from those changes:

1. (a)A 77% larger reduction in symptoms (5.3 − 3.0) ÷ 3.0 ≈ 77%
2. (b)0.8 of a day of recovered productivity a month per active user (0.4 − 0.2) days/week × 4 weeks = 0.8 days/month

   A month = 4 weeks, not 52/12 = 4.33 as in the 2026 study. That understates the 0.8 by about 8%.

That 0.8 of a day is about 22% of the loss from chapter 2 (0.8 ÷ 3.6). An active user therefore recovers about a fifth of the loss attributable to mental health.

Chapter 4 Both studies measure with the same two questionnaires

DASS-21 and WPAI are validated, widely used clinical questionnaires. We measure outcomes for clients after launch with the same set.

Chapter 5 At 1,000 employees the calculation gives 18–27 FTEs, and the plan pays back at one

The board calculator is six steps. Below, each of them for a company of 1,000 employees (HC).

1. (1)The loss in days HC × 3.6 × 1243,200 days of equivalent
2. (2)The value of the loss (1) × 355 PLN15.3M PLN a year≈ 171 FTEs
3. (3)Active users HC × the active share500 peopleat 50%
4. (4)The recovery in FTEs active × 0.8 ÷ 2218–27 FTEsat 50–75%
5. (5)The value of the recovery FTEs × 89,460 PLN1.6–2.4M PLNin year one
6. (6)The break-even point the plan's yearly cost ÷ 89,460 PLN93,600 ÷ 89,460 ≈ 1 FTE

50% is the share of active users we observe at clients – Mindgram operational data, not a study result. “Up to 36” in year two is the ceiling if every employee were active.

A different headcount, or a different cost of a working day? We will run it on your numbers

Chapter 6 Every assumption has a source and a direction it moves the result in

Of the eight assumptions, three understate the result (the cost of a working day, the inactive, working days), one can overstate it (the valuation method gives an upper bound), three cut both ways (the loss per person, the recovery per active user, the effect's persistence), and the result scales in direct proportion to the active share – hence the scenarios in figure 4.

Chapter 7 What these numbers do not prove

1. Self-report. Both studies are questionnaires completed by the respondents, with no HR data. Declared loss tends to be overstated, which is why we count only the excess above people with no symptoms.
2. No random assignment (2025). The active chose to use the platform themselves, so part of the difference may come from self-selection. We subtract the inactive group's improvement and speak of an association, not causal proof.
3. The comparison group is inactive users. The result compares using with not using within the same population. We did not study people with no access to support.
4. An internal study (2025), with no external review; we report group averages. The 2026 study was fielded by an independent research institute.
5. A cross-sectional study (2026). “Attributable” here means “the excess above people with no symptoms”. After adjusting for age, sex, sector and work model, 3.4 days remain, so the difference is not an artefact of the sample's structure.
6. Rebuilt weights, on the 2025 LFS working-population structure. Depending on the method the result is 3.6–3.75 days.
7. An online panel (CAWI) under-represents people with low digital skills.
8. The valuation is the upper bound of the human-capital method.
9. Extrapolation. We carry an effect measured between two waves onto a year and onto a new organisation. The real result depends on employee activity.