[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"article-body-predicting-misconduct-before-the-hotline-call":3},"\nStrip an ethics and compliance programme to its operating core and\nyou find the hotline: the channel where employees report what they\nhave seen, feeding the case management process that investigates it.\nThe field's best benchmarks are built on this machinery. NAVEX's\nannual analysis draws on millions of reports across thousands of\norganisations, and its 2026 edition describes case volumes, report\nrates and closure times, with closure times notably lengthening, in\nauthoritative detail.\n\nNotice, though, what every one of those numbers has in common: each\ndescribes something that already happened. A hotline report is filed\nafter the harassment, after the fraud began, after conditions in a\nbusiness unit deteriorated past someone's willingness to stay\nsilent. The function's primary instrument is a lagging indicator, and\nbenchmarking lagging indicators, however well, is measurement of\nharm, not prevention of it.\n\n## The signals that arrive earlier\n\nThe predictive information exists, and most of it sits in data the\nethics function already owns or can lawfully reach:\n\n- **Speak-up decay.** A team whose reporting volume drops to zero is\n  not a healthy team; healthy cultures produce a steady baseline of\n  questions and minor reports. Silence following a management\n  change, a restructuring or a missed target is a signal with\n  timing attached, visible in the hotline data itself, if anyone\n  models the baseline instead of celebrating the quiet.\n- **Retaliation risk markers.** What happened to the last three\n  people who reported in that unit: their ratings, their transfers,\n  their exits, relative to peers? Retaliation is both a harm and a\n  silencer of future signal, and its statistical shadow shows up in\n  HR data well before a retaliation claim is filed.\n- **Case-pattern drift.** Rising anonymity rates in one location,\n  substantiation rates diverging between units, repeat subjects\n  accumulating low-severity cases: individually routine, jointly a\n  forecast, when read across the portfolio rather than case by\n  case.\n- **The third-party blind spot.** Ethics teams increasingly own\n  third-party conduct risk while having audited only a fraction of\n  the portfolio; sector surveys find most programmes have assessed\n  well under half of the third parties they are accountable for.\n  The intermediary operating in a high-risk jurisdiction, with\n  ownership two layers deep and no completed diligence, is a\n  predictable incident with a name and an account number.\n\n> None of these signals requires new surveillance. They require\n> reading, jointly and statistically, the data the programme\n> already generates.\n\n## The regulator is already asking\n\nThe US Department of Justice's Evaluation of Corporate Compliance\nPrograms has, since its 2024 revision, asked prosecutors to probe\nwhether compliance functions have access to company data, whether\nthey use analytics on it, and how the company governs its own use\nof AI, including safeguards against misuse. The direction is\nunambiguous: a programme that cannot demonstrate it looks at data\nproactively is a programme whose adequacy will be argued about\nafter an incident, from a weak position.\n\nFor once, the defensive and the ambitious move are the same move. A\nprogramme that models speak-up decay, monitors retaliation\nmarkers, triages its third-party portfolio by predicted risk and\nroutes each signal to a named owner is simultaneously doing the\nthing regulators now ask about and the thing the function was\nalways for. The cases still get investigated; the hotline still\nruns. What changes is the tense the programme operates in.\n\nThe honest constraint deserves stating: predictions about people\ndemand more care than predictions about shipments. Signals should\ntarget units and portfolios, not individuals; thresholds should\ntrigger review, not accusation; and every model needs the same\npublished calibration and bias scrutiny you would demand before\ntrusting any consequential score. Prediction here is a way to\nallocate attention and support earlier, not a verdict engine.\n\nA test for your own programme: could you name, today, the three\nunits where speak-up volume has decayed fastest this year? The data\nto answer that is already in your case system; the Prophesee Ethics\nmodule reads it. [See what your case data is signalling](/contact).\n",1786833838013]