Wellbeing ROI: the model your CFO will actually accept

A wellbeing programme rarely dies in the People team. It dies in a finance review, when someone asks a simple question the business case cannot answer: what does this actually return, and how do you know? If your answer leans on a vendor slide promising a tidy multiple, the conversation is over. The credible move is to build an employee wellbeing ROI case the way a CFO builds any case, out of numbers you can source, assumptions you can defend, and losses you can show are already on the books.

That reframe is the whole game. Wellbeing is not a perk you are asking finance to fund. It is an avoided loss you are asking finance to stop absorbing. Poor mental health is already costing the organisation through absence, presenteeism, attrition, and the slow tax of disengagement. The question is not whether to spend, it is whether to keep paying for the problem or pay less to reduce it.

The World Health Organization estimates that depression and anxiety cost the global economy roughly US$1 trillion each year in lost productivity, with a similar order of magnitude of working days lost annually to these two conditions. Those are macro figures, but they describe something concrete happening inside your own headcount: people who are present but operating well below capacity, and people who quietly disengage before they leave.

Deloitte's work on the employer cost of poor mental health makes a consistent point at the organisation level: the largest line is usually not absence, it is presenteeism, the productivity lost when employees are at work but unwell. This matters for your model because a credible ROI case cannot rest on absenteeism days alone. Most of the cost is invisible on a leave register, which is exactly why finance tends to underestimate it.

In the Indian context the loss has specific texture. Long commutes, extended shift patterns, ESIC and PF cost structures tied to active headcount, and the cultural reluctance to declare a mental health day all mean a great deal of impairment shows up as presenteeism rather than recorded leave. Your starting number is not zero. It is large, and it is already being paid.

The number most often quoted in wellbeing decks comes from a WHO and Lancet analysis: scaled-up treatment of common mental disorders such as depression and anxiety is estimated to return roughly US$4 in improved health and productivity for every US$1 invested. It is a real, peer-reviewed estimate and a reasonable anchor. It is also frequently abused.

Two honesty rules keep it credible. First, the figure models scaled treatment of clinical conditions across whole populations, not the specific bundle of webinars and an app a vendor is selling. Treat it as directional evidence that investing in mental health pays back, not as a guaranteed return on your particular contract. Second, returns are not instant. The productivity and retention effects accrue over time, so any model that books the full benefit in month one is not a model, it is a wish.

A CFO will respect you more, not less, for saying this out loud. Presenting 4:1 as a benchmark from the literature, then building your own bottom-up estimate from your own headcount and cost data, is the difference between a business case and a brochure.

An honest model adds up three things you can actually estimate, then states what it leaves out. The first is returned productive days: a conservative assumption about how many impaired days are recovered when people get support earlier, valued at a fully loaded daily cost rather than base salary alone. Keep the recovery assumption low. A modest, defensible figure that survives scrutiny beats an aggressive one that collapses under it.

The second is reduced attrition. The cost of replacing an employee is commonly cited by SHRM and Gallup at somewhere between roughly one-half and two times annual salary once you count hiring, onboarding, lost productivity, and the load on remaining staff. In India you can add recruiter fees, notice-period overlap, and rehiring in tight talent markets. If a programme moves regretted attrition by even a point or two on a meaningful base, the avoided replacement cost is often the single largest line in the model.

The third is reduced absenteeism, the most visible but usually the smallest of the three. Value it from your own leave data rather than a benchmark. Then, crucially, write down what you are not claiming: brand reputation, recruiting advantage, and morale are real but hard to monetise, so leave them out of the headline number and mention them as upside. A model that under-claims is one finance can sign.

Return on wellbeing is not only a function of money spent, it is a function of whether the environment lets the spend work. Research on team performance points the same way. Google's Project Aristotle reported psychological safety, the shared belief that you can speak up without being punished, as the strongest differentiator of its most effective teams, a finding consistent with Amy Edmondson's earlier work on the concept. Where it is absent, people tend to hide problems until they become expensive.

Burnout is the cost signal to watch. The WHO's ICD-11 classifies burnout as an occupational phenomenon rather than a medical condition, and describes three dimensions: exhaustion, cynicism or mental distance from the job, and reduced professional efficacy. That last dimension is productivity loss by another name. Tracking these signals at an anonymised, aggregate level gives finance a leading indicator, not just a lagging spend report.

Whatever you measure, treat the data as sensitive. Under India's Digital Personal Data Protection Act 2023, wellbeing data carries obligations around consent, purpose limitation, and the rights of the data principal. Report to leadership in anonymised aggregates only. A programme that protects employee trust is also a programme that keeps producing usable signal, which is what makes the ROI durable rather than a one-off.

The strongest wellbeing case is the most modest one. Start from the loss the organisation is already carrying in presenteeism, attrition, and absence. Use the WHO and Lancet 4:1 figure as a benchmark from the literature, not a promise, then build your own bottom-up estimate from your own headcount and cost data. Keep recovery and retention assumptions conservative, state plainly what you are not claiming, and present a range rather than a single hero number.

Platforms like Aha can supply the aggregate, anonymised signal that keeps a model like this honest over time, but the discipline matters more than any tool. A CFO does not need to be convinced that wellbeing is worthwhile in the abstract. They need a model they can interrogate, trace, and defend. Give them that, and the conversation stops being whether to fund a perk and starts being how much avoided loss is worth paying to reduce.