Would it have warned you in time?
We ran the score across 18 years of market history, 2008 through 2025: 17,440 checks on 597 US dividend payers, covering the financial crisis, the COVID crash and every well-known dividend disaster in between. One thing to know about how it was built: the formula was locked using only pre-2020 data, and it performed just as well on everything that came after. Every number on this page is recalculated from that history when the site is built, not typed in.
How often each rating actually cut
Share of stocks in each rating that cut their dividend within the next 12 months, across the full 2008 to 2025 history.
| Rating | Cut within a year | In plain terms |
|---|---|---|
| Very Safe | 0.27% | almost never |
| Safe | 1.2% | about 1 in 100 |
| Borderline | 9.5% | about 1 in 10 |
| Unsafe | 27.7% | about 1 in 4 |
| Very Unsafe | 39.5% | about 2 in 5 |
Cuts it caught
The model was checked against the well-known dividend disasters: KHC 2019, T 2022, INTC 2023, VFC 2023, WPC 2023, LEG 2024, DOW 2025, GE 2017-18, KMI 2016, COP 2016, the COVID wave. Kraft Heinz sat deep in the danger ratings well before its 2019 cut. AT&T was flagged ahead of its 2022 reduction. Intel had already dropped to Borderline before announcing its cut in 2023.
The misses, honestly
A handful of companies rated Safe still cut. Nearly all of them were healthy businesses that suspended dividends in the first weeks of COVID, when revenue went to zero overnight. No financial model catches that. The other notable miss was Intel, which deteriorated faster than its next scheduled rescore. None of the misses came from the Very Safe rating.
We publish the misses because a safety score you can't audit is just marketing. If you want to check our work, the full method and the complete scored history are open.
Get the next warning in your inbox, free
This model rechecks every stock weekly. When one gets downgraded, subscribers hear first. No spam, unsubscribe anytime.
For the statistically minded: the model's ranking skill (AUC) was 0.828 on the years it was built on and 0.811 on 2020 to 2025, which it never saw. A score of 1.0 is perfect and 0.5 is a coin flip. Barely losing accuracy on unseen years is the point.