Ask a room of HR directors what a falling number of workplace conduct reports means and you will get a pause, then a careful answer. Most know there is a trap in the question. Fewer can say precisely what it is.
The trap is this. A reporting figure is produced jointly by two things: what happened, and how safe people felt saying so. From inside the reporting system, those two inputs cannot be separated. A quiet quarter is consistent with a healthier organisation and equally consistent with a collapse in confidence. The chart looks identical either way.
This is not a marginal technical caveat. It is the central limitation of an entire category of organisational data, and decisions are being made on that data every week.
The providers have said so themselves
What makes this a live issue rather than an academic one is that the companies selling reporting platforms have started saying it out loud.
Culture Shift, one of the largest UK providers of workplace reporting software, surveyed more than a thousand UK employees at the beginning of 2026 and found a pervasive fear of speaking up about bullying and harassment. Separate research they have published identifies a widening confidence gap between junior employees and senior leaders on whether speaking up is safe, with senior people markedly more confident. Their chief executive has argued in the trade press that organisations should stop treating misconduct as a reporting problem and start addressing the trust issues that prevent people speaking up in the first place.
That is a notably honest position for a company whose product is reporting software. It is also correct, and it has an implication that the sector has been slower to follow through on.
Following the diagnosis to its conclusion
If fear is what suppresses reporting, then the reporting dataset consists of disclosures from people who overcame that fear or never felt it. Everyone whose fear is held is, by definition, absent.
Improving the reporting process, better anonymity, cleaner intake, faster triage, more visible channels, expands the group that overcomes the fear. This is worth doing and it genuinely helps people. It does not convert the person who has assessed their situation and concluded that disclosure carries more risk than silence, because that person’s reluctance is a judgement about consequences rather than a difficulty with the form.
And the composition of that absent group is not random. The confidence gap points directly at it. The people least likely to appear in the data are those with the least power, the shortest tenure and the most to lose. The people reading the data are those most insulated from the dynamic being measured. An organisation reviewing its speak-up analytics is looking at a picture drawn disproportionately by the people least exposed to the problem.
The distance between prevalence and reporting
The scale of the gap is easier to see when you set the two figures side by side.
Estimates cited across the sector suggest that around 40 percent of women and 18 percent of men have experienced some form of unwanted sexual behaviour in the workplace. Formal reporting sits nowhere close to those figures in any organisation, however good its systems.
That distance is not process failure. It is a population, and it is a large one. Every person in it made a calculation and reached a conclusion. Better forms do not change the calculation, because the calculation is about what happens afterwards.
Which is why organisations that have invested seriously in speak-up infrastructure are still, with some regularity, blindsided. They did not have bad data. They had data that answered a question adjacent to the one they thought they were asking.
What a board should actually ask
Three questions change the quality of the conversation considerably.
If these numbers halved next quarter, would we know whether that was good news? If the answer is no, and it usually is, the figures cannot support the conclusions currently being drawn from them.
Who is missing from this dataset, and is it the same people every time? Cross-referencing reporting rates against tenure, seniority and department frequently reveals silences that are more informative than any of the reports.
What do we know that never became a report? Exit interviews, informal escalations, patterns raised and dropped, the things everyone in a team knows and nobody has written down. This information exists in most organisations. It is rarely brought into contact with the formal data.
Where the other measurement comes from
The structural answer is that measurement cannot depend entirely on someone choosing to come forward. Something is needed that does not sit downstream of the disclosure decision.
Documented workplace communication is such a source. Emails, messages and internal correspondence exist as a by-product of ordinary work, generated whether or not anyone ever complains. They are not created by a reporting decision and so are not filtered by one. Analysed for patterns rather than read as individual incidents, they can indicate where a dynamic is present in a part of an organisation that has filed nothing at all.
This is what CultureScan does, and it is worth being precise, because the market around it is not always careful with language. It does not attempt to determine how employees are feeling. It analyses documented communication for structural patterns, and every finding can be traced back to the specific material supporting it.
That distinction is not merely a design preference. European regulators have drawn a version of the same line. The EU AI Act prohibits the use of AI to infer emotions from biometric data, meaning faces, voices and similar signals, in workplace settings, and the reasoning given in the legislation is that such inference lacks a sufficient scientific basis to be relied upon where power is unequal. Analysis of what was actually written sits outside that prohibition, and for a reason worth noticing: the evidence can be inspected by the people it concerns.
Used well, this does not replace a reporting platform. It answers a question the reporting platform structurally cannot: what is happening in the parts of the organisation that have gone quiet.
The uncomfortable version
Most organisations would rather have a reporting problem than a trust problem, because a reporting problem can be solved with procurement.
The evidence, including evidence published by the reporting providers themselves, points at the trust problem. And a trust problem cannot be measured by a system that only records the behaviour of people who already trust it enough to use it.
A quiet quarter might be good news. It is worth being able to tell.
