Hetty Green’s principles, and what this app took from them

This app is named after Henrietta Howland Robinson Green, born in New Bedford, Massachusetts in 1834 and dead in 1916. She spent her adult life investing — railroad bonds and shares, government debt, mortgages, city real estate — and by the end of it she was widely described as the richest woman in America. The financial press of her day called her the Witch of Wall Street. She was not a witch. She was a lender holding cash at moments when almost nobody had any, which looks close enough to sorcery if you happen to be the one who needs the money. (We spell it with one t. She had two.)

Naming a research tool after an investor invites a fair question: which parts of what she did can actually be built into software, and which parts were simply her? What follows is our answer, and it is deliberately unflattering to us in places.

First, what we are not going to do

Most of what circulates about Hetty Green is anecdote. She was mocked in print for four decades, and the stories that survived were selected for how well they made her look absurd — the same black dress, the unheated rooms, the haggling over trivial sums. Some of it is probably true. Little of it is documented in a way that would survive a footnote.

The quotations are worse. Lines credited to her circulate widely, in several different wordings, with no source anyone can point to. So we are not going to quote her. Everything below describes a method in our own words, drawn from what she is reliably known to have done rather than from what she is supposed to have said. An app whose entire argument is that it does not invent figures should not introduce itself by inventing a founder’s voice.

What she is reliably known to have done

She was the buyer when everyone else had to sell

Her career spans the panics of 1873, 1893 and 1907, and the pattern in every account is the same: she went in liquid, and she bought or lent while prices were being set by people who had no choice about selling. Accounts of the 1907 panic have her lending to New York City itself, at a point when the banks had stopped.

This is the least mystical idea in investing and the hardest one to actually carry out, because it requires having spent the preceding years doing something that felt like a mistake at the time.

The cash was the strategy, not the leftovers

Being the buyer in a panic is not a decision made during the panic. It is the consequence of holding reserves and avoiding debt through all the years when both look like errors — when the cash earns little and the leverage would have paid handsomely. She held large reserves and borrowed sparingly. Those reserves were not caution left over after investing. They were the position.

The corollary matters more than the rule. In a panic, whoever is forced to sell is funding whoever is not. Her whole method was to make certain she was never on the paying side of that trade.

She valued things she could go and look at

She favoured assets with something concrete underneath: a mortgage on a specific building, a bond secured on a specific railroad, land she had walked over. She is reported to have inspected collateral herself, which in the 1880s meant getting on a train.

The discipline underneath is not really about property. It is a preference for claims that can be verified over claims that have to be accepted. A bond secured on track you have seen is a different proposition from a story about a growing industry, even when the arithmetic looks similar.

She preferred to be paid contractually

Much of what she held paid interest or rent — returns that were written down and owed, rather than dependent on a more optimistic buyer arriving later. That does not make bonds better than equities. It makes the source of the return legible, which is a different claim and frequently a more useful one.

She did her own work and ignored the enthusiasm

Every account agrees here: she read the documents, kept her own books, and formed her own view. She was conspicuously unmoved by tips, promoters and the prevailing mood. This is the part of her method most obviously worth copying and least obviously easy, because doing the work properly on one company takes an evening, and there are five hundred of them.

The thrift was about the one return she controlled

The stories about her economising are told as eccentricity. Read as method, they are a recognition that costs are certain while returns are not. Every dollar of expense is a known subtraction from a return that can only be estimated, so she minimised the part that was known. Whether she carried it too far is a question about her life rather than about her arithmetic.

What this app took from it

Six ideas, and what each one actually became. These describe behaviour the app already has, not intentions for it.

1. Demand a large discount, mechanically

The screener estimates a conservative value range for every company in the S&P 500 and the Dow 30, then derives an entry range from the low end of that estimate: a third of it up to a half of it. Put plainly, the app will not describe a price as attractive until it sits somewhere between a half and two thirds below the cautious estimate of what the business is worth. That is a severe margin of safety by design, and it is why the deeply undervalued list is usually short and sometimes empty.

Illustrative and hypothetical: if the conservative low estimate for a company came out at $90 a share, the entry range shown would be $30 to $45. The point of that arithmetic is that most prices do not qualify. A screen that flatters the market is not doing anything.

2. Publish the assumptions so they can be argued with

Value is estimated two independent ways — a decade of reported earnings priced at a fixed sector earnings multiple, and the balance sheet priced at a fixed sector price-to-book — and both sets of multiples are published constants rather than tuned inputs. Two people running the app on the same company get the same range, and can see precisely which assumption they disagree about. This is the desk-bound version of going to look at the collateral: you cannot inspect a railroad from a browser, but you can inspect a model, and only if somebody shows it to you.

3. Use tests designed to disqualify

Graham’s seven defensive tests run on every company: size, leverage, earnings stability, dividend record, earnings growth, price to earnings, price to book. They are not blended into a score. Each one passes or fails, and the failures are shown as failures. A checklist whose job is to find reasons not to buy is far closer to how she is described as having worked than any rating out of ten would be.

4. Check what management said against what they filed

The app reads a company’s annual report and its earnings call and reports the places the two do not line up — on margins, on demand, on guidance, on capital allocation, on risks disclosed in one document and described rather differently in the other. This is the feature with no analogue in her lifetime, and it follows straight from her method: she is nowhere described as having taken anybody’s word for anything.

5. Show a blank rather than an estimate

Where a figure is genuinely missing from the filings, the app shows that it is missing. It does not interpolate it, infer it, or quietly substitute a sector average, because a fabricated input produces an output that looks exactly like a real one. A number is only checkable if the gaps around it are still visible.

6. Watch the forced sellers

The spin-off calendar exists for the reason she watched panics. When a company hands a division to its own shareholders, shares arrive in accounts that never chose them, and are often sold for reasons that have nothing to do with the business — an index mandate, an awkward position size, plain indifference. Joel Greenblatt built a body of work on that mechanic decades after her death. It is the same underlying observation: the most dependable source of a mispriced asset is a seller who has no choice.

She was liquid when being liquid was unfashionable. Almost everything else followed from that one decision.

Where the comparison stops

It would be easy to end on the flattering half, so here is the rest of it.

She had advantages that do not transfer. She invested in a market with far less public information, far fewer people running the same analysis, and no requirement that a company disclose much of anything on a schedule. A real part of her edge was being one of the very few doing the work at all. That particular edge has been competed away, and any tool implying otherwise is selling something.

Her results are not a benchmark we are claiming. We publish no track record, because we do not have one that would mean anything — this is a research tool rather than a fund, and the decisions, along with their consequences, belong to whoever uses it. Nothing here should be read as suggesting that a checklist derived from her method produces an outcome resembling hers.

And the concentration that built her fortune is the part we would most caution against copying. By any modern standard she was undiversified, and it worked for her. That is a fact about one person over one lifetime, not a general recommendation, and it is not what this app is built to encourage.

What survives translation, then, is narrower than the legend and rather more useful: insist on a wide discount, prefer what you can check to what you are told, read both documents, and keep the gaps visible. None of that requires being extraordinary. It mostly requires not skipping steps, which is the part software is genuinely good for.