You are looking at 1,831 US companies over $1bn, scored from their own filings and ranked inside their own sector.
This page is about driving the app. For what the measures mean and how they are calculated, see DOCS.
This app reads the accounts of every large American company and sorts them, industry by industry, so the interesting ones come to the top.
Every company listed in America has to file its full accounts with the regulator, by law, and those filings are public. Revenue, profit, debt, cash — all of it.
About 1,800 companies worth more than a billion dollars. Nobody could read that by hand, which is the only reason this app exists.
Is it growing, does it make money, does the money turn into cash, can it survive a bad year, is it cheap, is anyone inside buying — and so on.
Never against the whole market. A supermarket and a software company are not comparable, and pretending otherwise would just list the software companies.
Every number can be opened to reveal the formula, the actual figures, and a link to the page of the filing it came from. You are meant to check it.
What it does not do: predict anything, tell you what to buy, or know anything about the future. There is not a single forecast in it. It reads what has already been filed and arranges it so you can think.
If a word on this page is unfamiliar, it is in the glossary below in plain English. That is not a footnote — it is the point. You cannot operate this app without the vocabulary, and the vocabulary is about twenty words.
The eight tabs are eight different questions. Nothing is “the best company” in the abstract — a name that tops COMPOUNDERS will usually look expensive under OWNER, and that is not a contradiction. Choose the tab that matches what you are actually looking for.
Rankings are computed inside each sector, because a 20% margin means one thing in software and another in grocery. The list is grouped by sector for exactly that reason. Untick group by sector if you want a flat list.
It is the share of the scoring that had data behind it. A 95 at 45% judged is a thin result wearing a confident number — worth less than an 80 at 90%.
Click any row and the card opens: the eight scores, then what every measure found, in sentences. Nothing is asserted without the finding behind it.
Every figure with its formula in words, the formula with the actual numbers, and for each input the XBRL tag it came from and a link to the filing on EDGAR. Two clicks from a ratio to the document.
The star saves to the server, so the list is the same on every device. Anything you star is included in the nightly enrichment pass whatever its score.
TEASER PDF on the card gives a one-page summary — the case, what to watch, the numbers, the filed horizon, insider activity and the news. Printable, and readable in greyscale.
One company, read from top to bottom. No finance background assumed.
You can lend to it, or you can own part of it. They are completely different deals.
Lending is a bond. The company promises fixed interest and your money back on a fixed date. If it survives, you get exactly what was promised and not a penny more. If it goes bust, you are near the front of the queue.
Owning is a share. You own a slice of everything the business earns, forever, with no promises at all. If it does well you keep doing well; if it fails you are last in the queue and usually get nothing.
The Oracle only covers shares. If what you want is a fixed income with a known end date, this is the wrong tool and it will not tell you otherwise.
Cash in your account — a dividend. The company posts you part of its profit, usually every quarter.
Your slice becoming worth more — either the business grows, or the company buys back its own shares so the same business is split among fewer owners.
MANHATTAN ASSOCIATES INC pays you the second way only. It reports no dividend of any kind in its filings — no dividend-per-share figure and no dividends-paid figure. Instead it spent $315m in its last full year buying its own shares back. That is why the teaser shows share count -5.1%: fewer owners of the same company, so each remaining slice quietly grows.
So if you are looking for income to live on, this is not that kind of share. If you are looking for something that compounds and returns cash by shrinking itself, it is.
Almost every reason to own a company falls into one of these, and each of the tabs is one of them:
Read straight off its own scores, inside Business Services (incl. Software):
Put in one sentence: an excellent business, priced richly, paying no dividend, handing cash back by buying its own shares. That is a growth-and-quality share, not an income share — and whether the price is worth it is a judgement the Oracle deliberately does not make for you.
Everything below is in the frame on the right. Tap a number there as you read about it here.
It is printed in amber, never in red. A high price is a fact about the market, not a fault in the business — and the two must not look the same. A wonderful company can be a poor purchase on the day, which is the whole reason the OWNER tab exists.
None of this is advice, and there are no forecasts anywhere in it. Every figure is something a company filed, or a delayed price from Interactive Brokers. The point of showing the arithmetic and linking the filings is that you are meant to check it.
Every number on the teaser, in plain words, with MANH’s own figure beside it.
The tag on the right of each is the Oracle’s reading of that figure. strong, fair and weak are judgements against a threshold; expensive is a price fact and not a criticism; per share means the figure was measured but is not good or bad on its own; not measured means the company did not file what was needed.
Of the money the owners have put in and left in, how much the company earns back each year.
Above about 15% is good. Be careful though: borrowing heavily makes this number look better without the business being better, which is why return on capital employed sits next to it.
net income ÷ shareholders' equity = $220M ÷ $315M
The same idea, but against everything the company owns — factories, stock, cash, the lot.
Above about 10% is good. A business that needs very little equipment can score highly here; a railway never will, and that is not a fault.
net income ÷ total assets = $220M ÷ $839M
What the business earns on ALL the money tied up in it, borrowed and invested together.
The single most useful measure of quality on the page. Above about 12% is good; above 25% is rare. Unlike return on equity it cannot be flattered by debt.
operating income ÷ (equity + long-term debt) = $280M ÷ ($315M + $0)
How many dollars of sales the company squeezes out of each dollar of things it owns.
High means it does a lot with little. Low is normal for software and for anything that owns heavy machinery, so read it against the industry rather than on its own.
revenue ÷ total assets = $1.08B ÷ $839M
What is left out of each dollar of sales after ALL the running costs — wages, rent, marketing, everything.
Above 20% is strong. This is the cleanest view of whether the day-to-day business makes money.
operating income ÷ revenue = $280M ÷ $1.08B
What is finally left after running costs, interest and tax.
Above 10% is healthy. It can be lifted by one-off gains, which is why the operating margin is shown beside it.
net income ÷ revenue = $220M ÷ $1.08B
Operating profit with the accounting charge for wear-and-tear added back — roughly the cash the operations throw off before the bills.
Useful for comparing companies that own a lot of equipment. It is NOT profit: the wear-and-tear is real and eventually has to be paid for.
(operating income + depreciation & amortisation) ÷ revenue = ($280M + $6M) ÷ $1.08B
Out of each dollar of sales, how much actual cash is left after keeping the business running and its equipment up to date.
The hardest number on the page to fake. Profit can be shaped by accounting choices; cash arriving in the bank cannot.
(operating cash flow − capital expenditure) ÷ revenue = ($389M − $15M) ÷ $1.08B
How many years of the company's current profit you are paying for the shares.
Under about 20 is modest, above that you are paying for growth that has not happened yet. A loss-making company has none, which is why it sometimes shows n/a.
market cap ÷ net income = $12.02B ÷ $220M
The same thing against sales instead of profit.
Used when profits are small or absent. Under about 5 is modest.
market cap ÷ revenue = $12.02B ÷ $1.08B
The price against what the accounts say the company owns after its debts.
Under 1 means the market values it below its own books. Very high is normal for a business whose value is people and software rather than buildings.
market cap ÷ shareholders' equity = $12.02B ÷ $315M
How many years of the cash the company actually generates you are paying for the shares.
The strictest of the four, because it uses cash rather than profit. Under about 15 is modest.
market cap ÷ free cash flow = $12.02B ÷ $374M
What the WHOLE company costs — the shares plus the debt you inherit, less the cash that comes with it — against its rough operating cash.
This is what a buyer of the entire business would look at. Under about 12 is modest.
(market cap + debt − cash & investments) ÷ EBITDA = ($12.02B + $0 − $329M) ÷ $286M
The profit divided by the number of shares: your slice of a year's earnings.
Not good or bad on its own — $3 of earnings could be cheap or dear depending on the price. Watch whether it GROWS.
net income ÷ weighted average diluted shares = $220M ÷ 61M shares
The sales divided by the number of shares.
Its direction over years is the point: sales rising while this stays flat means new shares were issued and you gained nothing.
revenue ÷ weighted average diluted shares = $1.08B ÷ 61M shares
What the accounts say the company owns, after debts, per share.
A floor of sorts for asset-heavy businesses. Nearly meaningless for one whose value is its staff and its software.
equity ÷ weighted average diluted shares = $315M ÷ 61M shares
Whether the number of shares grew or shrank over the window.
Shrinking is good for you: the same business divided among fewer owners, so your slice grows without you doing anything. Growing means your slice was diluted.
change in weighted average diluted shares over the window = 64M → 61M shares
What the company can turn into cash within a year, against what it owes within a year.
Above 1 means it can cover the next year's bills without borrowing. Below 1 is not automatically fatal but it needs a reason.
current assets ÷ current liabilities = $583M ÷ $456M
The same, but ignoring unsold stock — because stock is only worth something if somebody buys it.
Above 1 is comfortable. For a company with no stock at all it equals the current ratio, which is normal and not a sign of strength.
(current assets − inventory) ÷ current liabilities = ($583M − $0) ÷ $456M
Every line ends with the arithmetic: the formula in words, then the same formula with the company’s actual numbers in it. The SHOW THE NUMBERS button on any card goes one step further and gives the XBRL tag and a link to the filing each input came from.
Growth, margins, cash conversion, a solid balance sheet. Says nothing about the price — that is what OWNER is for.
Book value and cash rather than earnings multiples. Expect unglamorous companies here; that is the point.
Measured against the company's OWN historic margin, not against its industry.
Open-market Form 4 purchases only. Grants and tax withholding are not decisions.
The only view that weighs the business AND the price. Start here if QUALITY keeps handing you expensive names.
Position in the 26-week price range. A fact about where the price has been — not a prediction.
The only growth you receive. Catches companies that grew revenue by issuing shares.
The profitability floor is the whole point: it keeps companies that are simply failing out of the list.
A low score on a measure the current tab does not weight is not a fault. Manhattan Associates tops QUALITY at 100 while scoring 0 on recovery, because it sits at its own peak margin — perfect for a compounder, disqualifying for a turnaround. The readable view only warns you about measures the tab you are on actually cares about.
The default. A ranked list where each entry carries its reasoning in sentences. Use it to decide what is worth opening.
Every measure at once as a heat grid, sortable by any column. Use it when comparing a specific metric across many companies is the job.
Every view is addressable, so you can bookmark a filtered list or send somebody straight to one company.
oracle.html?q=MSFTopens the explorer filtered to a searchoracle.html#favoritesopens filtered to your starred namesoracle.html?view=gridopens straight into the dense heat gridoracle.html?card=MSFTopens a company card directlyoracle.html?workings=MSFTopens the arithmetic panel directlyteasers/MSFT.pdfthe teaser PDF, if that company has oneEvery one of these is easy to do and none of them is stupid. They are the predictable ones, so they are worth knowing before you make them.
It means “this matched what the tab measures”. QUALITY does not look at the price at all, so its top name can be a terrible purchase today. Use OWNER when you want the price counted.
A 96 on 45% of the measures is a thin result wearing a confident number. Always glance at it before you get excited.
It is amber, not red, on purpose. Being expensive is a fact about the market, not a weakness in the business.
A 6% margin is poor for software and excellent for a grocer. That is why the list is grouped by industry and the percentile is inside it.
A dash means the company did not file that figure. The Oracle never scores an absence as zero, and the card lists exactly what was missing.
On INSIDER those are opposite facts. “Not read” means the filings could not be fetched.
Every price here is delayed and refreshed once a night. It is for deciding what to read, never what to pay.
There are none anywhere in this app. Everything is either something a company filed or a price. If you want a view on the future, that is your job, and the workings exist so you can form one.
Roughly in order. Nobody needs all of it on the first day, and the last group is the part that actually matters.
51 words, grouped by the question you are trying to answer rather than alphabetically — because somebody who does not know the word “margin” also does not know to look under M.
No definition here uses another word that also needs defining. Where one genuinely depends on another it says so, rather than sending you in a circle.
One slice of ownership in a company. Own a share of a company worth $100 with 100 shares, and you own $1 of it plus your slice of everything it earns from now on. Nobody promises you anything.
A loan you make to a company. It pays you fixed interest and hands your money back on an agreed date. Safer than a share, and that is the whole trade: you cannot lose as much, and you cannot gain as much.
In the app: The Oracle does not cover bonds at all.
Cash the company sends you, usually four times a year, out of its profit. A share that pays one is often called an income share.
In the app: There is no dividend measure in the Oracle yet, so it cannot rank income shares.
Instead of sending you cash, the company spends money buying its own shares and destroying them. Same business, fewer owners, so your slice gets bigger without you doing anything.
In the app: Shown as SHARE COUNT CHANGE. A negative number is a buyback.
The opposite. The company issues new shares, so the same business is split among more owners and your slice shrinks. Often used to pay staff or to buy something.
In the app: Caught by the PER SHARE measures: revenue can rise while revenue per share does not.
What the whole company costs at today's price: the share price times the number of shares. A $50 share with 100 million shares means a $5 billion company.
In the app: The Oracle only covers companies above $1 billion.
How many shares exist. You need it to turn a share price into the price of the whole company.
All the money customers paid the company. The top line. It says nothing about whether any of it was kept.
What is left of the revenue after every cost: making the thing, wages, rent, interest on debt, and tax. The bottom line.
In the app: The lens called PROFIT. Papi's word for it is utilidades.
Profit as a percentage of revenue. Sell $100 and keep $20, and that is a 20% margin. It tells you how much of each sale the company actually gets to keep.
What is left after paying only for the thing that was sold, before wages and rent. A software company keeps almost everything; a supermarket keeps very little.
What is left after all the day-to-day running costs, but before interest and tax. The cleanest view of whether the actual business works.
Money genuinely moving in and out of the bank. Different from profit, because profit includes things that have not been paid yet.
The cash left over after the company has kept its equipment and premises up to date. This is the money it can actually do something with — pay you, pay down debt, or buy something.
In the app: The lens called CASH.
Operating profit with the accounting charge for wear-and-tear added back. A rough measure of cash from operations, used to compare companies that own a lot of equipment.
In the app: It is NOT profit. The wear-and-tear is real and eventually has to be paid for.
Short for capital expenditure: money spent on equipment, buildings and machinery — the things needed to keep operating.
Everything the company owns: cash, buildings, machines, stock on the shelves, money customers owe it.
Everything the company owes: loans, bills, wages not yet paid.
Assets minus liabilities. What would be left for the owners if the company sold everything and paid everyone off — according to its own accounts.
Money borrowed, which has to be paid back with interest whether or not business is good. Debt is not bad in itself; it is dangerous when earnings wobble and the interest still falls due.
In the app: The lens called SOLV, short for solvency.
All the money tied up in running the business: what the owners put in plus what was borrowed.
In the app: The denominator of return on capital employed, the best single quality measure on the page.
Accountants split things by timing. Current means within a year, non-current means later than that.
In the app: The split of contracted revenue into current and non-current is what the HORIZON reading is built from.
Money a customer has already handed over for something not yet delivered. A gym paid for a year up front owes you eleven months of gym.
In the app: How the Oracle tells a long-term business from a short-term one.
Work already sold but not yet delivered or counted. A builder with three years of signed contracts has three years of backlog.
In the app: Filed as remaining performance obligation. It answers “how much of next year is already won”.
Any “price divided by something”. It turns a price into a number you can compare between companies of different sizes.
In the app: Shown under WHAT THE MARKET CHARGES.
How many years of the company's current profit you are paying for the shares. A P/E of 20 means twenty years of today's profit.
In the app: Under about 20 is modest. Above it you are paying for growth that has not happened yet.
The price against what the accounts say the company owns after debts. Under 1 means the market values it below its own books.
In the app: Very high is normal where the value is people and software rather than buildings.
What the whole company costs a buyer: the shares, plus the debt they inherit, less the cash that comes with it.
P/E turned upside down, written as a percentage. A P/E of 20 is a 5% earnings yield, which is easier to compare against a savings rate.
The highest and lowest the share price has been over the past year. Knowing the price is $50 tells you nothing; knowing it has been between $30 and $55 tells you a lot.
In the app: The TREND measure is where in that range the price sits today.
A price a few minutes old rather than live. Fine for deciding what to read; useless for deciding what to pay.
In the app: Every price in the Oracle is delayed.
One thing the Oracle measures — revenue, profit, cash, returns, debt and so on. Each gives a mark out of 100 and writes down why.
The same lenses added up with different importance. QUALITY cares about margins and cash; OVERLOOKED cares about being cheap. Eight tabs, eight questions.
The lenses blended, out of 100, for whichever tab you are on. Change tab and it changes. It is not one truth about the company.
Where that score sits among the company's OWN industry. 90 means only 10% of its industry scored higher.
In the app: Comparing a supermarket to a software company on margin is meaningless, which is why everything is ranked inside its industry.
How much of the scoring had real data behind it. A 95 built on half the measures is thinner than an 80 built on all of them.
In the app: The single most useful number for not fooling yourself.
STRONG, FAIR and WEAK are judgements against a threshold. EXPENSIVE is a fact about the price and not a criticism. PER SHARE means measured but not good or bad on its own. NOT MEASURED means the company did not file what was needed.
The panel that opens when you click a company: its scores, and what every measure found, in sentences.
The SHOW THE NUMBERS button. Every figure with its formula, the actual numbers, the exact tag it came from and a link to the filing.
In the app: Built so you can check the app rather than trust it.
A one-page PDF for a company: what it does, why it screens, the numbers, and its news. Printable and shareable.
The slow extras — news, insider filings, the business description. Too expensive to do for all 1,800, so it runs for the leaders of each industry plus your favourites.
In the app: For anything else, open its card and press ENRICH.
Whether the company's revenue is already contracted years ahead or has to be won again next year. Taken from what it filed, not from opinion.
The public copy of the app. It has no server behind it, so it is frozen at the moment it was built and cannot save favourites.
In the app: It says so in a banner across the top.
The US financial regulator. Every listed American company must file its accounts with it, truthfully, by law.
The big annual report a company files. Item 1 of it describes the business; Item 1A lists what could go wrong.
In the app: The “what the company does” paragraph on a teaser is quoted from Item 1.
The SEC's public library where all those filings live. Free, and the workings panel links straight into it.
The tagged, machine-readable version of those accounts. It is why 1,800 companies can be read automatically instead of by hand.
In the app: A “tag” in the workings panel is the exact label the company filed a number under.
The form a director or officer must file after buying or selling their own company's shares.
In the app: Only open-market PURCHASES count as conviction. Shares handed over as pay, or held back to cover a tax bill, are not somebody choosing to buy.
Somebody who runs the company — a director or a senior officer. Not a criminal term; it just means they know it from the inside.
The industry code the company itself files under. Older and blunter than modern schemes, but it comes from the company rather than from a third party.
In the app: The Oracle ranks inside these.
A company's own twelve-month accounting year, which need not match the calendar. Some end in June, some in January.
In the app: It is why a figure can look a year out of date and be current.
Treat every score as a question, not an answer. The app is built so you can always get from a number to the filing it came from in two clicks — and the only reason to build it that way is that you are meant to check it.
Research and education only. Not investment advice.