Questions, answered plainly

Three things, in this order. It shows you a company's numbers with the source of each one attached: the formula, the inputs, the statement they came from, its fiscal year and its filing date. It lets you write down why you own the company, in your own words, together with the numeric conditions that have to stay true for that reason to hold. Then it re-tests those conditions every night against newly filed data and emails you the first night one of them fails.

When nothing breaks, you hear nothing.

People who pick their own stocks and expect to hold them for years. The typical subscriber researches between three and ten US-listed companies a month and reads the filings.

Traders, and anyone who wants the software to make the decision. There is no buy list, no rating and no price target of our own, because a number we invented would sit on the page with nothing behind it. Where an analyst consensus target appears it is labelled as someone else's opinion, so you always know whose judgment you are reading.

No. It gives you the filed figure, the formula behind it, the statement it came from and the date it was filed, so the decision is yours and you can say exactly why you made it. Then it re-tests the reasons you wrote down every night and writes to you the day one of them stops being true.

Nothing here is advice, a recommendation or a forecast, and that is the point: a tip you cannot check is worth nothing the day it stops working.

More detail: Risk disclaimer.

A spreadsheet is as current as the last evening you spent on it. This reads the new filings every night, holds the reasons you bought beside the numbers that carry them, and writes to you the day one of those numbers breaks. It also keeps the source statement behind every figure, so you are checking the filing rather than a cell you typed six months ago.

No. Every score expands into the tests behind it, and a verified figure into the statement it came from, so the product shows its working instead of asking you to trust a rating. You do need to be willing to read a balance sheet.

No. There is one plan and one price, and a 14-day trial of the whole product. The trial takes a card at the start and charges it on day 15 unless you cancel first. A free tier would mean either gating the monitoring service, which is the product, or running it for free, which is the expensive part.

More detail: Pricing.

The United States: NYSE, NASDAQ and AMEX. One filing regime is the reason a figure here can open into the statement behind it. Foreign companies listed here are covered as listings, and where we hold their filed accounts their figures are read in the currency they report in. Many file on 20-F or 40-F forms we do not yet ingest, so they carry vendor figures rather than a source trail, and their pages say so. A listing that trades only in Europe or Asia is absent, not thin.

Two counts, and both are printed live at the top of this page: how many companies you can look up, and how many have their filed accounts on disk. The second is the set we can prove anything about, so it is the one quoted wherever a source trail is promised.

Neither is ever typed into the page. The verified count climbs every night as filings are ingested, and a number written into the copy would be stale by morning.

No, and which group a company is in is the first thing its page tells you. Where the filed accounts are on disk the full promise holds: a figure traced to its statement, recomputed from source, and monitored for a breach. Where they are not, you are never asked to trust the number anyway.

Every listing falls into one of three tiers. Verified means the statements are on file. Metrics only means an indicator value exists with no statement behind it, so it cannot be sourced, recomputed or monitored. Listed only means the company is in the tradable universe with no usable fundamentals at all.

If a company you follow is in a thinner group, send us the ticker and we will look at pulling its filings.

Yes, nightly. Companies that filed since the previous pass are ingested every night, so the verified count at the top of this page climbs as the work lands. No date is promised for any particular company.

If a company you follow is thin, tell us the ticker. It goes to the front of the queue.

No. They are excluded from the equity surfaces and from the ranking pools, because a percentile rank is only meaningful against comparable operating businesses. A fund has no Piotroski score to compute.

Prices are stamped at the previous US close, and every page that prints one carries that date. For a holding period measured in years, a settled close is the number that matters, so nothing here is real-time by design. A decision that turns on the last few seconds needs a trading terminal.

Fundamentals move on a different clock. They come from filed accounts, which change when a company files rather than when the market does.

Because they are different facts, and conflating them would let a stale quote pass as fresh. Recomputing an indicator updates the computation timestamp without refreshing a single price.

The accounts each company files, standardised into comparable statements. Every statement carries its fiscal year, its period, its filing date and its reported currency, which is what makes the source trail on each figure possible.

More detail: Methodology.

A failed or partial fetch never overwrites data we already trust. Where two sources disagree, the figure carries a mismatch badge rather than being silently reconciled. Where a value is missing, the product prints n/a, never a zero and never an estimate dressed up as a fact.

Up to ten fiscal years on screen, where the filings exist, and fewer where they do not. The research page shows the years it has, so a short history reads as a short history rather than as a flat line. Nothing is padded, interpolated or extended.

There is a hard floor worth knowing about: machine-readable filings only begin when structured reporting was phased in, around 2009 to 2011. A company that has traded for fifty years therefore has roughly fifteen years of tagged data, and a company that listed last year has one.

A cross-sectional percentile rank against the scored universe, not a rating and not a price target. A score of 80 means the company ranked above roughly four fifths of the universe on the underlying indicators at the time it was computed. Extremes are trimmed at the 1st and 99th percentiles so a single outlier cannot distort the scale.

More detail: Methodology.

Five pillars, weighted: Value at 35 per cent, Quality at 30, Health at 15, Growth at 12 and Risk at 8. A pillar score is the plain average of the available indicator percentiles inside it, and the overall score is the weighted sum of the five. The weights on the methodology page are the ones the engine uses.

More detail: Methodology.

A nine-point checklist of accounting improvement published by Joseph Piotroski in 2000: profitability, cash flow, leverage, liquidity and operating efficiency, each test scoring one or nothing. It is shown as published, with the nine tests expandable, and it is not a valuation measure.

More detail: Methodology.

A distress model from 1968 that blends five ratios into one figure and reads in zones: above 2.99 distress is unlikely on this model, 1.81 to 2.99 is inconclusive, below 1.81 is elevated risk. It was fitted on US manufacturers, so treat it as a prompt to look at the balance sheet rather than as a verdict.

More detail: Methodology.

An earnings-manipulation screen from 1999 built out of eight indices. It flags the statistical signature of aggressive accounting, which tells you which notes to read first; it is not a finding of fraud. It needs two consecutive years of statements, so it is shown where those exist and marked absent where they do not.

More detail: Methodology.

n/a means the figure is missing. Zero is a number and missing is not: writing zeros into gaps would drag every average and every percentile rank towards a value nobody reported. A gap is shown as a gap, on every surface, including inside the formula view.

n/m means not meaningful. The usual case is return on equity where equity is negative: the arithmetic produces a number, the number is nonsense, so it is flagged rather than printed as if it were informative.

Yes, on every figure. Expanding one opens the formula, the input values, the statement they were read from, the fiscal year and period, the filing date and the reported currency. If a figure cannot show that trail, the product tells you it cannot rather than showing it anyway.

It runs a deliberately simple two-stage discounted cash flow: one growth rate for the explicit stage, one discount rate you enter yourself, and a Gordon growth perpetuity for the terminal value. It shows you what share of the result sits in that terminal term, which is usually a large one, and it saves versions so you can see how your own assumptions moved over time.

The discount rate is yours to argue for, not ours to compute for you, and there is no fade path or scenario weighting to hide a number behind. Every input is one you can defend six months later, which is what saved versions are for: they show which assumption moved. The simplifications are listed in full on the methodology page.

The fair value it produces is not a target price. It is the arithmetic consequence of the assumptions you typed in, and changing the discount rate by a point moves it a long way, which is the most useful thing a DCF teaches.

More detail: Methodology.

16 of them, each a published value discipline expressed as filters, and each naming the book or paper it comes from. Where a published rule uses a figure we do not hold, the card says so instead of quietly substituting another one.

Every result carries its forensic scores, so a cheap company that is also distressed cannot hide inside a list of cheap companies.

Only with care. The percentile ranks are cross-sectional against the whole scored universe rather than sector-neutral, so a capital-intensive utility and a software company are ranked on the same scale. Compare across sectors using the pillar breakdown and the underlying indicators, not the headline number.

Your reasons for owning a company, written in your own words, plus the numeric conditions that have to stay true for those reasons to hold. The words are for you. The numbers are what the service can honestly re-test.

One condition and one threshold, phrased as the statement that keeps your thesis alive. "Return on invested capital stays at or above 20 per cent" is a guardrail. It is breached the night that statement stops being true.

About five minutes for the first one. The builder suggests three guardrails for the company you are looking at, pre-filled with its current values, so the starting point is an edit rather than a blank page.

Every night at 23:30 UTC, after the indicator snapshot has run, so the check sees the freshest values available.

Within a day of the data reflecting it, rather than whenever you next reread the filings. It is not a price feed and does not race one: the market will have moved before the filing reaches us. What it ends is holding a broken thesis for months without knowing.

No. A guardrail fires once, on the transition from holding to broken. While it stays broken the check runs and stays silent. If the value comes back inside your threshold, a recovery note is written to your inbox with no email, and the guardrail is re-armed for the next time.

Nothing fires, and the state is left untouched. A data gap is a fact about our pipeline, not evidence about your company, and an alert that cannot tell the two apart is worse than no alert.

Yes. Standalone alerts on a single indicator are swept every thirty minutes and re-armed daily at 01:00 UTC, so a condition that stays true does not mail you repeatedly.

Three things, and each can be switched off on its own in Settings: an alert when one of your lines is crossed, a note when a company you follow files new numbers, and one weekly brief on Sunday covering what moved on your watchlist and what your theses did. Every alert email carries a one-click unsubscribe.

No. It monitors the figures you chose, computed from filed statements. A headline is not a breach, and a product that treated it as one would be a signals service.

Then write that part as text, where it belongs. Your bull and bear case sit alongside the numeric guardrails, and a breach alert points you back at both. Only the numbers get re-tested nightly, because only the numbers can be re-tested honestly.

No. There is no analytics script, no tag manager, no advertising pixel and no session recorder anywhere in the site. The only inline script is the one that picks your colour theme before the first paint.

There is one cookie, and it holds the value 1 so the server knows to send a signed-in visitor to the product rather than the marketing page. Three keys sit in local storage: the token that keeps you signed in, your theme choice, and the tickers you looked at recently. That is the complete list.

More detail: Privacy policy.

On a machine in the founder's office in Spain, reachable only through a Cloudflare tunnel, with the PostgreSQL database on the same machine. The origin accepts no inbound connections of its own. A nightly snapshot is copied to a second machine we control.

In transit, yes: the snapshot moves over SSH to a second machine we control. At rest, not yet: the nightly dump is compressed but not encrypted. The origin accepts no inbound connections of its own and sits behind a tunnel; encryption at rest is still to come.

Passwords are hashed with bcrypt at twelve rounds and never stored in any other form. Refresh tokens are stored only as SHA-256 hashes, are single-use, rotate on every refresh, and are revoked on sign-out. Repeated failed sign-ins lock the account temporarily.

Stripe for payments, and card details never reach us. Resend for transactional email and Brevo for product email. Cloudflare for DNS and the tunnel. Slack, which receives your email address and display name on account and billing events so the founder sees them.

The privacy policy lists each one with its purpose and the transfer safeguards.

More detail: Privacy policy.

Only if you tell us. Theses, guardrails, watchlists, notes, positions and cost bases are all optional user-supplied content, they belong to you, and the licence we take over them covers hosting, running the nightly checks, delivering alerts and taking backups.

More detail: Terms of service.

$29 a month, or $228 a year. One plan, no feature gates, no seat tiers. Tax is added at checkout where the law requires it.

There is no launch rate, no founding-member price and no countdown anywhere on this site.

More detail: Pricing.

A 14-day trial of the entire product, and yes, it needs a card. Nothing is charged until day 15, when the plan you picked begins. Cancel in one click before then and you pay nothing at all. One trial per account.

The card is taken at the start so the trial ends by beginning your subscription rather than by locking you out mid-thought.

On the Stripe checkout page, in the field marked add promotion code. It applies to the first charge and to every renewal the code covers, and the amount it took off is shown on the invoice.

Codes come from us directly. There is none hidden on this site to be found.

Your subscription begins at the price you picked and the card on file is charged. There is no gap and no lockout.

If you cancelled during the 14 days, nothing is charged and access runs to the end of the trial. Your theses, saved valuations and watchlists are kept as you left them, and only the nightly monitoring stops.

One click, on the billing section of Settings, inside the app: one question about why, no retention call. The renewal stops and you keep access until the end of the period you already paid for, or to the end of the trial if you are still inside it.

A first payment is refundable in full for thirty days, with no form to fill in. That is more generous than the fourteen-day statutory withdrawal right, and it is written into the Terms rather than being a policy we could quietly change.

More detail: Terms of service.

Money behind the one job that matters: if a guardrail you were monitoring broke and we did not email you within a day of the data reflecting it, we refund your last two months. Ask and it is paid, by a person. It covers our silence, not the data's: where the figure never arrived, arrived late, was restated afterwards, or was missing on the night of the check, the check was right to say nothing.

More detail: Terms of service.

Stripe, through Stripe Checkout. Card numbers never touch our servers and we never see them. The trial is a Stripe trial, so the date you are charged is held by Stripe rather than by us.

Subscription state is reconciled against Stripe every ten minutes and again the moment you return from checkout, so a lost webhook cannot leave your account locked out of something you have paid for.

Write to us and it is changed by hand. There is no self-serve edit yet.

Ask at the address in the privacy policy. Your subscription is cancelled as part of the request, so no renewal is taken while it is processed, and the deletion is completed within thirty days. It is not self-serve yet.

More detail: Privacy policy.

Screener results and portfolio data export as CSV from inside the app, and you can request a copy of everything held about you under the privacy policy.

More detail: Privacy policy.

Use the reset link on the sign-in page. It sends a single-use link by email, and completing a reset revokes every existing session on the account.

There is no team plan and none is planned. The product monitors one person's reasons for owning something, and sharing a login would mean sharing a thesis board, which is the part that should not be shared.

The app carries a service log: what ran, when it ran, and what it found. The point of paying for standing work is being able to check that the work happened.

Report it. Every research page carries a report-a-data-issue link, and it goes to a person. Open the figure's source trail first if you can, because knowing which statement and which fiscal year it came from turns a vague report into a fixable one.

Prices are stamped at the previous US close, and the date sits next to the figure. Nothing here is a live quote.

Because that listing is in the metrics-only tier: an indicator value exists but the statement behind it was never archived, so there is nothing to show you. Those values cannot be sourced, recomputed or monitored, and the product marks them rather than dressing them up.

It is listed outside the United States, or it is a fund, warrant or preferred, which are excluded on purpose. Tell us the ticker and it goes on the list.

Four things to check, in order. The value may have been missing on the night of the check, in which case nothing fires by design. The guardrail may already have fired and be waiting to re-arm. That email type may be switched off in Settings. And it may be in your spam folder, in which case marking it as not spam helps the next one arrive.