Actionable insights — The ultimate stock screener?

1. Encode the quality checklist as eight hard thresholds, with the averaging window chosen per metric

1. Growth, measured per share: 5-yr revenue CAGR >5% and 5-yr diluted EPS CAGR >7% — diluted so share issuance (SBC, convertibles) can't manufacture growth.
2. Margins on multi-year averages: gross margin (5-yr) >40% for pricing power, net margin (5-yr) >10% for resilience, FCF margin (3-yr) >10% so earnings are backed by cash.
3. Capital intensity: CapEx / revenue <5%.
4. Capital allocation, two ways: ROIC (3-yr) >15% — above any plausible cost of capital — and ROCE (3-yr) >20%.
5. Require all eight. The point is elimination, not ranking: "Every step makes the list of possible investments smaller."

• Averages hide a trend: a 5-yr gross margin above 40% can be a falling 55% → 38%. Look at the series for anything that passes.
• A 5-yr EPS CAGR from a depressed base year (a 2020-21 trough) will flatter cyclicals and recoveries.

2. Scope the universe before filtering — exclude what the metrics can't judge and what you can't

1. Exclude industries whose economics invert the metrics (banks: deposits are "debt", interest is revenue — ROIC and gross margin mean nothing).
2. Exclude countries outside your circle of competence or your portfolio's rules.
3. Pick exchanges deliberately: including OTC markets brings in foreign lines (useful) and illiquid microcaps (usually not).
4. Note that country filters key on domicile/HQ as the data vendor records it, not where the business operates.

• A screen whose exclusions disagree with the portfolio's own rules (here the developed-countries-only policy): add the stricter rule as a filter, or the shortlist will keep surfacing names you can't buy.
• Add a minimum market cap / liquidity floor if the universe includes OTC.

3. Sort the survivors three ways to see what kind of quality you've caught

1. Sort the passing list by revenue growth, by EPS growth and by ROIC, and read the top ten of each.
2. Names at the top of all three (here Nvidia, PDD) are the growth-and-returns outliers; names only on ROIC are the mature cash machines (Rightmove, Games Workshop); names only on EPS may be margin or buyback stories.
3. Take the shortlist for work from across the sorts, not just the leaders.

• Extreme EPS CAGRs (IHG 235%) almost always mean a tiny or negative base year — check before treating it as growth.
• If you publish leaders, quote the actual sort order; the post's lists skip IHG, Kaspi and Booking.

4. For each finalist, write "how does the company make money?" in two sentences, then name the moat

1. State the business in plain words (what is sold, to whom, how often).
2. Name the single mechanism that keeps the numbers high: monopoly technology, embedded distribution, legal/contractual lock-in, scarcity brand, network effect.
3. Pull one operating chart that evidences the mechanism (volumes, AUM, margins) rather than the share price.
4. Only then move to valuation — the screen and the moat say nothing about price.

• A moat story that ignores cyclicality in the end market (ASML's customers order in waves; 2024 revenue was nearly flat).

5. Verify each filter actually bit — check the output column, not the settings box

1. After running, look at the column for every criterion in the results: its minimum should sit at or just above the threshold.
2. If a column is uniform (all 0) or blank, the metric is unpopulated or the unit is wrong (0.05 vs 5%); fix it and rerun.
3. Re-count survivors with that filter on and off to see how much work it is doing.

• Screens that return suspiciously familiar names: a non-binding filter often looks fine because the other seven still find good companies.

6. Scan the survivors for data artefacts before naming any

1. Flag any gross margin of exactly 100% — the company reports no cost-of-sales line, so the gross-margin filter was passed automatically (use operating margin instead).
2. Flag ROCE or ROE in the hundreds or thousands — usually negative or near-zero equity from buybacks, not superior economics.
3. Flag market caps below your liquidity floor and names from excluded geographies that slipped through on domicile.

• Artefacts cluster in exactly the business models the house likes (payment networks, platforms, buyback-heavy compounders) — they pass for the wrong reason, which matters when the screen is later tightened.