Opinions are cheap. These are computed.
Four engines run nightly across the whole universe. None of them is an AI guessing — every number has a formula behind it, and every formula is locked by a test that runs on each change.
Four engines run nightly across the whole universe. None of them is an AI guessing — every number has a formula behind it, and every formula is locked by tests that run on each change.
High marks on the three quality pillars and a fair — not cheap — price. The sizing rule puts this in the top conviction band.
Suggested weight: 8–12% of the portfolio
In a test on US filers 2009–2022, the fundamentals behind the score (85 of its 100 points) ranked the next three years’ peer-relative returns modestly well; its valuation inputs, label cut-offs, Canadian listings and sizing bands are untested.
The Deep Score
Quality and value in one number, 0–100.
Every pillar is ranked against sector, size and business-class peers, so a utility isn't judged by a semiconductor's margins — and the headline is where that total ranks across the whole universe, recomputed nightly. Price is counted once: valuation is 15 of the 100 points, and the quality pillars that set the discount rate leave it out, so a cheap stock cannot talk its own fair value up.
The Inflection Engine
Six weighted signals on the rate of change.
Share-count discipline, margin expansion, operating leverage, improving returns, earnings surprise and cash-flow inflection, each ranked against sector peers — with five more shown, no weight, growth acceleration and insider buying among them. Built to surface companies inflecting into quality before the trailing averages catch up — orthogonal to the Deep Score on purpose, so the two disagreeing tells you something.
The signal engine
27 rules, read against a market regime classified from 5 inputs.
VIX, the US 10- and 2-year Treasury yields, the high-yield credit spread and the S&P 500 against its 200-day average classify the regime; the rules then read every holding against it, and the currency rules also read the CAD/USD trend. Rates, oil and inflation are shown beside it as context. SELL outranks TRIM outranks ADD, enforced — so you are never handed two instructions that contradict each other.
Fair value
Fourteen models, blended, with a confidence.
A DCF you can drive with sliders, a reverse DCF that solves for the growth the market is already implying, and a Bull/Base/Bear range built only from the company's own normalised cash flow. When the models disagree, the confidence drops and says so.
One number, and the reasons behind it.
The Deep Score
Every company gets one number out of 100 for how it stands among the names we scanned.
Each pillar is ranked against sector, size and business-class peers, so a utility is not judged by a semiconductor's margins; the headline then ranks that standing across the whole universe.
Price counted once
Valuation is 15 of the 100 points — and the quality half that drives the discount rate excludes it.
So a cheap stock cannot talk its own fair value up: price reaches the appraisal through the cash flows and the multiple, never twice. A three-year decline is flagged beside the score too.
The value-trap guard
A weak business is never promoted just for being cheap.
A trusted margin of safety can move a position one notch at most — and never off the floor.
Conviction position sizing
Shows the band a transparent rule assigns to a score, as a reference for your own sizing.
The score sets the band: avoid at 0%, high conviction at 8–12%, never above a single-name limit you declare. The bands, and the label cut-offs that set them, have not been tested against later returns.
Score trend
A small chart showing how the name's standing has moved across recent scans.
The Deep Score is a rank within each night's scanned universe, so the line moves when the field changes as well as when the company does — it is a series of standings, not a measurement of the business over time.
Earnings quality
Checks whether the reported profit is backed by real cash.
Free cash flow ÷ net income, badged STRONG through RED FLAG below 0.6.
Fourteen answers, and an honest confidence.
Most screeners ship one DCF and present it as a fact. This one runs an ensemble and tells you when they disagree.
Fourteen models, blended
Fourteen different ways of valuing a company, combined into one number.
Three DCF variants, six multiples, four asset-based models and a dividend model.
It says when it doesn't know
When the models disagree, the confidence drops and the app says so.
Only a HIGH or MEDIUM confidence fair value is allowed to move a position-sizing tier.
Reverse DCF
Shows what the market is already assuming about a company at today's price.
Solves for the implied growth rate. Above 25% is flagged as near-perfect execution priced in.
Bull, base and bear
Three honest what-if prices instead of one made-up target.
Built only from the company's own normalised cash flow, averaged across profitable years so a cyclical peak can't inflate the bear case.
Peer comparison
Put two or three companies side by side.
Twelve metrics, a five-year return chart, and a ranked verdict.
Smart money
Shows who else owns it and whether the insiders are buying.
180 days of insider activity, and a warning when institutional ownership passes 80%.
The score tells you what a business is. This tells you what it’s becoming.
Six weighted signals on the rate of change — share dilution, margin expansion, operating leverage, improving returns, earnings surprise and cash-flow inflection — each ranked against sector peers, with five more shown as facts with no weight. Built to be orthogonal to the Deep Score, so the two disagreeing is information rather than a contradiction.
6 of 6 weighted signals strong · confidence chip reports how many had data
All six weighted signals read strong against sector peers at once, and the Deep Score has not caught up yet — the trailing averages are still carrying three bad years. This is the case the engine exists to find.
We built a probability. Then we shipped a rank instead.
The original engine put a percentage on it — the odds a company becomes a long-run winner. The data underneath could rank candidates honestly, but it could not certify a calibrated probability at the depth we actually have. So the percentage came out and the ordering stayed. It surfaces the same names; it just refuses to tell you a number it can’t stand behind. A tool that shows its work has to be willing to delete its best-looking output.
Before the trailing averages catch up.
Emerging Compounders
Finds companies that are just starting to get great, before the trailing numbers show it.
Six weighted signals on the rate of change, each ranked against sector peers, plus five shown, no weight — banded Inflecting 80+, Emerging 65–79, Building 50–64.
A confidence chip
Says how much data it actually had to work with.
How many of the six weighted signals had data — fewer than four gets no score — and heavy dilution caps the label regardless.
ETF screening
The same quality scoring, for funds.
Cost, size, yield and 1/3/5/10-year returns, recomputed from adjusted price history rather than the feed's own figures.
Ranking that follows your objective
The fund ranking changes depending on whether you want growth or income.
An aggressive-growth profile de-weights yield from 15% to 5% and lifts long-term performance from 35% to 50%.
Thesis guards
Tripwires on the business, not the price.
Score, margin, debt and growth conditions — and they only fire after two consecutive scans.
A record of what you looked at
Remembers every company you researched and what it cost then.
Price and score at first view, and the change since.