Apple fundamental analysis: 4 of 8 idle

Run eight numeric guardrails backwards over the ten fiscal years of Apple 10-K filings this product holds, read 2026-09-07, and four of the eight never fire once. Two would have broken in most years on file, and two behave the way a guardrail is supposed to.

A thesis with no numbers in it cannot be checked, so it cannot be wrong, so it never gets revised. The fix is to attach numeric guardrails: statements that must stay true for the reason you bought to hold.

The failure mode of that fix is less obvious. A guardrail that could not plausibly be crossed is decoration, and a guardrail that is already crossed is noise. Both feel like rigour while you are writing them, and both are visible in about a second if you run them backwards over what the company actually filed.

That is a fundamental analysis run in reverse: instead of asking what the company looks like now, it asks which of your rules the filed record would already have broken. Here are the eight guardrails, each run over the fiscal years this product holds as 10-K filings.

Apple Inc., from GET /api/public/research/index, read 2026-09-07. The fiscal years are 2016 through 2025 and every value behind them comes from a 10-K this product holds.

Four of the eight are decoration

The top four never fired. On this company, over this decade, return on equity never dropped below 15 percent, operating margin never dropped below 15, net margin never dropped below 10, and free cash flow was never negative.

A guardrail like that is not wrong. It is simply not information. If you write it into a thesis you have added a line that will never wake you up, and the absence of an alert from it means the same thing on the day you wrote it as it will in five years.

The bottom two are the mirror image. Debt to equity has been above 1.0 in eight of the ten years and the current ratio has been below 2.0 in all ten. A guardrail there does not monitor the company. It disagrees with it. Writing "debt to equity stays below 1.0" into a thesis on this company is not a condition for the thesis holding, it is a statement that the thesis was already broken when you wrote it.

The 1.0 line is Graham's, from chapter 14 of The Intelligent Investor. It is a real threshold with a real source, and it is the wrong threshold for this company, which has spent a decade buying its own shares back with borrowed money. A guardrail at 3.0 would have held throughout and would have fired if the policy changed. That is a different and more useful sentence.

The two that behaved like guardrails

Gross margin at or above 40 percent. Broken in 2016 through 2020, held from 2021 onward.

The same threshold is a dead guardrail if you wrote the thesis in 2016 and a live one if you wrote it in 2021. Nothing about the number changed. What changed is where the series sits relative to it, and that is the only thing that determines whether a guardrail can do work.

Revenue growth of at least 5 percent. Broken in 2019, 2023 and 2024. Three fires in nine years is the shape of a guardrail that is neither decoration nor disagreement: it holds most of the time and it does not hold always.

Nine years, not ten. Growth compares two consecutive filed years, so the first year on file has nothing to compare against and is not evaluated. The tool reports nine rather than quietly treating the first year as a pass. The same arithmetic caps every streak read out of the filed archive, though not the dividend streak, which is read from a longer record and reaches 55.

The tension this creates, stated plainly

There is an obvious way to abuse the above: look at the history, then set every threshold just below where the series has been. You would produce eight guardrails that have never fired and a thesis that cannot break, which is exactly what we started out trying to avoid.

The threshold has to be the number at which you would actually change your mind, and the retro-test does not choose it for you. What it does is tell you what you have written. If the number you would genuinely act on has never been crossed in a decade, that is worth knowing before you rely on it as a monitoring device. If it has been crossed in eight years out of ten, that is worth knowing too, and it usually means the guardrail is measuring something about the industry rather than something about the company.

The useful sequence is: write the threshold you mean, run it backwards, and then decide whether what came back is the alert system you thought you were building. Not: run it backwards and pick a number that looks good.

What the alert then does

Once a guardrail is written, it is re-checked against the newest filed data and fires once, on the transition from holding to broken. It does not repeat while the condition stays broken, because the second email carries no information the first did not.

That behaviour is what makes the composition of the list above matter. Four guardrails that never fire cost nothing to hold and add nothing. Two that are permanently broken would either fire immediately and then go silent, or would have to be excluded from the alert entirely. Two that fire occasionally are the reason the mechanism exists.

What this leaves out

One company, eight rules, ten fiscal years. That is an illustration of a method and not a sample of anything. Nothing here is a view on Apple or on any other company, and none of these thresholds is recommended.

The years covered are the fiscal years this product holds as EDGAR filings for this company, 2016 through 2025. Accounts filed before a company began tagging its statements in XBRL are not machine-readable and are outside every one of these tests.

Five of the eight thresholds are conventions chosen as lines to test rather than published standards. Two, the 1.0 debt to equity and the 2.0 current ratio, come from Graham's chapter 14. One, free cash flow above zero, is definitional. The product states the source of each threshold next to it, and none of them is a boundary this product estimated.

Only eight figures can be run backwards over filings at all. Anything with a share price in it, and most composite scores, cannot be rebuilt for an old year with the same formula as the live value, so no honest test of them exists.

Past filings do not indicate future results. A guardrail that held for ten years is not a prediction that it will hold for an eleventh.