Miss Risk · beta
Commitments at risk of being missed
Every company promises dates. Some of those dates are already slipping before the deadline arrives. Miss Risk flags them early, from what the company itself has said and how it has delivered before.
Back-test: trained only on deadlines before 2022, then scored 2,299 later commitments it had never seen. The riskiest fifth was missed 54% of the time; the safest fifth 12%. This is one time-split test on our own data, not independent validation: it uses outcomes whether or not they are confirmed, some inputs may not have been public on the scoring date where the evidence date is unknown, and it has not yet been compared with simple baselines such as event type and time to deadline. A frozen, pre-registered validation with calibration and those baselines will be published before Miss Risk is described as more than a beta research signal. It is not a trading signal or investment advice.
334 open commitments at high risk across 177 companies, 277 elevated
Which companies and which commitments are at high Miss Risk, the signals behind each rating, and an alert the day a commitment you follow moves into high risk, are part of KeptIndex Pro. Here is where the risk sits today, by sector.
| Sector | Open commitments scored | High risk | Elevated | Share at high risk |
|---|---|---|---|---|
| Utilities | 52 | 32 | 7 | 62% |
| Consumer Discretionary | 86 | 30 | 22 | 35% |
| Materials | 135 | 46 | 34 | 34% |
| Health Care | 368 | 112 | 104 | 30% |
| Information Technology | 78 | 21 | 20 | 27% |
| Energy | 104 | 27 | 34 | 26% |
| Financials | 103 | 23 | 19 | 22% |
| Communication Services | 41 | 9 | 3 | 22% |
| Industrials | 115 | 23 | 19 | 20% |
| Real Estate | 44 | 7 | 6 | 16% |
| Consumer Staples | 30 | 4 | 6 | 13% |
| Other | 9 | 0 | 3 | 0% |
See which commitments are at risk, before the deadline
KeptIndex Pro shows the Miss Risk rating and its signals for every open commitment at the companies you follow, emails you the day one moves into high risk, and sends the full list each week as a spreadsheet.
How Miss Risk works
Experimental, for research use. Miss Risk is designed to use only facts on the record at the time of scoring. Our own validation has found that this is not yet guaranteed: about a third of checked records lack a known evidence date, so some historical tests may include information that was not yet public, and the model has not yet been compared with a simple baseline. Treat the bands as a research signal, not a basis for decisions. It uses:
- The company’s own record: how often its earlier commitments were delivered on time, counted only once each outcome was public, and whether its most recent one was missed.
- What it has said since: whether the company has already restated this commitment, pushed the deadline later or softened the wording (for example from “will” to “aims to”).
- The commitment itself: its kind (a new plant is harder than a filing), how precisely the deadline was stated, how far ahead it was set, and whether delivery depends on a condition.
A model weighs these signals using 7,244 commitments whose outcomes are known. Raw model probabilities are never shown, because they can be over-confident. Instead each commitment is placed in a band, and the page shows how often commitments in that band were actually missed in the past, measured on records the model was not trained on.
Miss Risk is a statistical estimate, not a judgment about any company and not an outcome. Only commitments that passed our checks are scored, and the score never changes a recorded outcome or a Kept Score. If a company believes a commitment is on track, it can claim its page and add a public update.