The slow way, made survivable Chapter 5 of 23 Svenska

The map · Part One · The method · Chapter 5 of 23

Why a name is refused, and why that is the normal outcome

Most companies that reach the reader are sent away. The gates are designed to say no, a failed gate stays failed, and refusing to judge a business you cannot judge is a real answer. About 9 minutes.

By the time a name reaches a person, it has survived a coarse screen. It is cheap-ish, it makes money, it has not obviously broken. The temptation is to think the hard part is done. It is not. The reading is where most names are refused, and refusal is the expected result, not a disappointment.

The gate that does most of the refusing: why did the price fall?

The screen found companies whose price fell. The reader's first job is to name the reason, and put it in one of four boxes.

  • A: the mood changed. Nothing about the business changed; the story people tell about it did.
  • B: the whole sector was sold. Money left an industry and took this company with it.
  • C: a one-off cost with a number on it. A fine, a write-down, a factory fire. It happened once and its size is known.
  • D: something is actually broken. A product that no longer sells, a customer that left, a regulator that has changed the rules for good.

A, B and C pass. D fails, and the rule is stricter than it looks: if the reader cannot say with confidence which box it is, it is D.11Framework §3 gate 2: four disconnect classes; 'if you cannot classify confidently, it is D.' A real name was dropped on class D on 2026-09-08. Not knowing why a price fell is not neutral. It is the most common way to buy a broken business.

Worked example: NAME A

NAME A's price is 35 percent below last year's high, so it passed filter 1. The reader opens the last four reports. Revenue is flat. The fall began the week the company's largest customer, a fifth of its sales, announced it would make the product itself from next year. That is a lost customer that is not coming back: box D. NAME A fails gate 2 and leaves. Its cheapness was never the question.

A name failing a gateNAME A passes gate 1 with a tick and reaches gate 2, where the fall is classed as lasting damage and fails with a cross. An arrow leads down to a box reading out: WATCH-GATED or DROPPED. A dashed line from that box back toward NAME A ends in a cross, labelled: a lower price later does not bring it back. only a named new fact reopens a failed gate: a filing,a guidance change, a profit warning. Never a level on a chart.NAME Agate 1how far hasthe price fallen?gate 2: why did it fall?class D:lasting damageout: WATCH-GATED or DROPPEDa lower price later does not bring it back:the gate failed on something other than price
NAME A fell far enough to pass gate 1. Gate 2 asks why it fell; the answer is lasting damage, and that is a fail. It leaves. A cheaper price later does not bring it back; only a named new fact can.

The other gates

A name that passes the "why" gate is then tested on its accounts over the last two years: debt within a stated multiple of earnings, interest covered at least five times by operating profit, spare cash generated in at least six of the last eight quarters, no auditor's doubt about the company continuing, no change of auditor, no restated accounts. Then it needs a dated event within ninety days that could make the market look again: a results date, a product decision, a ruling. "Eventually the market will realise" is not an event.22Framework §3 gates 3 and 5. Gate 4, a valuation discount against peers, is recorded as DATA MISSING for every name because the peer figures it needs have never been assembled; it is not deleted, and returns when they exist (ruling E99).

Then there are seven hard kills: things that end the reading on their own, whatever else looks good. Revenue falling two quarters running. Margins shrinking while revenue is flat. Two cuts to the company's own forecast in a year. Three missed earnings in two years. Debt above a stated multiple or close to breaking a lending agreement. Both the chief executive and the finance chief leaving within a year. Stock or unpaid invoices piling up faster than sales.33Framework §4.2. 'Any one invalidates.'

Why a failed gate is final

The natural way, which is wrong

NAME A failed on a lost customer at a price of 60. Six months later the price is 40. Surely at 40 it is worth another look? The price has done half the work.

The rule, and why

The price was never the problem. NAME A failed on a fact about its business, and at 40 that fact is still true. A name that failed on something other than price is parked as WATCH-GATED, and the only thing that can bring it back is a named new fact: a new filing, a changed forecast, a warning. A price level cannot reopen it, because a price alert would wake it into the same gate it is already known to fail.

This is written into the method as a ruling with a number, and it is the reason the tool keeps two kinds of waiting list apart. A name that is merely too expensive waits for a price. A name that failed a gate waits for a fact.44Ruling E27 splits the old single 'watch' status into WATCH-PRICED and WATCH-GATED; the retired spelling is refused by the code with a message pointing at the ruling.

The same finality applies to the rules themselves. When two limbs of a gate were found to measure nothing useful, they were removed, and the ruling that removed them says "not widened, not softened, removed". A rule is either in force or it has been retired by a dated decision. There is no third state where it is bent for a name that nearly passed.55Ruling E30, 2026-08-25, deleting two limbs of gate 3 that measured the volatility of reported figures rather than the quality of the business, having changed no verdict across five names.

The circle: refusing what you cannot judge

Some businesses are refused before any gate, not because they are bad but because the method cannot be run on them honestly. The method needs the owner to write down a believable range for how the business will grow (chapter 8). For some businesses that is not a judgement anyone can make.

  • A company whose revenue is a world price it does not set, multiplied by a volume: a miner, an oil producer, a bulk shipper. Its bad case is not slower growth but a halved price, which is a different business, not a lower number in the same arithmetic.66Ruling E96, 2026-08-31: 'This is a limit of the METHOD, and it is not a judgement about the industry.' Sixteen industry patterns; regulated utilities, contracted output and semiconductors expressly excluded from the rule.
  • A drug developer, whose growth is a guess about approvals the owner has no way to assess.77Ruling E51.

Buffett's phrase for this is the circle of competence: "The size of that circle is not very important; knowing its boundaries, however, is vital."88Buffett, 1996 letter to shareholders. What this project adds is that the boundary is written down as a rule and enforced by the screen, so that a spectacularly cheap miner cannot tempt anyone into pretending the method applies. Such a name is never watched at all, not even on its price, because there is no level it could reach that would mean anything.99Ruling E97: a name outside the circle 'is NEVER WATCHED AT ALL, not even on its drawdown'.

Refusing a business you cannot judge is not a cop-out. It is the only answer that does not require pretending.

The exact rules

Framework §3 (gates), §4.2 (hard kills), §4.4 (conviction score). Rulings E27 (the two waiting lists), E30 (deletion of two limbs), E51 and E96 (the circle), E97 (never watched), E99 (gate 4 as DATA MISSING). Code: the status list in vss/rules.py, the circle lists under config/.

Sources for this chapter

  1. Buffett, 1996 letter to shareholders (circle of competence)
    https://www.berkshirehathaway.com/letters/1996.html
  2. Circle of competence, Wikipedia
    https://en.wikipedia.org/wiki/Circle_of_competence
  3. Gates, hard kills and the circle rulings — read from this repository: reference/FRAMEWORK.md §3–§4; reference/FRAMEWORK-EDITS.md E27, E30, E51, E96, E97, E99

The slow way, made survivable. Built 2026-09-13 from the repository as it stood that day. NAME A and NAME B are fictional; no company, price or holding appears on this site. Not financial advice: one private investor's method, shared as it is.