Actionable insights — HEICO: A 186-page Deep Dive

1. Score a business on fixed metrics — then read the sub-scores, never the average

1. Run every candidate through the same fifteen headings — business model, management, competitive advantage, industry, risks, balance sheet, capital intensity, capital allocation, profitability, stock-based compensation, historical growth, outlook, valuation, owner's-earnings evolution, historical value creation — and score each out of ten with the evidence written beside it.
2. Compute the total as the plain average, but treat it as a filing label rather than a verdict.
3. Read the distribution. A respectable total can hide a catastrophic single score, because averaging is precisely what dilutes it.
4. Nominate in advance which sub-scores are vetoes rather than contributors. Valuation and main risks are the natural candidates: no amount of business quality repairs an unpayable price or an unacceptable single point of failure.
5. Keep the failing evidence in the sheet rather than arguing it away, so the next reader sees the same trade-off.

• Scoring drift — the same analyst grading a favourite business half a point higher across every heading.
• Metrics that double-count one strength (capital intensity and capital allocation both rewarding the same capital-light model).

2. Use a reverse DCF as a veto: compare the implied growth with the delivered growth

1. Instead of forecasting cash flows, solve backwards: at today's price and a required return, what annual free-cash-flow growth rate is being assumed?
2. Put that number beside the company's own realised growth over five and ten years.
3. If the implied rate exceeds the best period the business has ever delivered, the price is not a forecast — it is a bet on an improvement nobody has demonstrated.
4. Say so numerically in the write-up, so the pass is falsifiable and the level that would change it is implicit.
5. Keep the name on a follow list with the yield as the tracking metric, and revisit when the price moves rather than when the story does.

• A reverse DCF whose required return assumption is doing the work; state it explicitly (10% here, per this source's standard).
• Serial acquirers whose historical growth was acquisition-funded — the implied rate must be achievable with the same deal cadence and the same balance sheet.

3. Treat "quality everyone already knows about" as a structural obstacle, not a compliment

1. Ask whether your view of the business differs from consensus at all. If it does not, the only remaining source of return is the price.
2. Check whether the shares have ever traded at a discount to their own history. A name that never de-rates offers no entry point, only a queue.
3. Distinguish this from a de-rated compounder, where fear has created a gap between quality and price — that is the setup this source normally buys.
4. Decide explicitly which of the two you are looking at before doing any more work, because the second half of the analysis differs completely.

• Waiting forever for a de-rating that a genuinely scarce asset never delivers; a follow list needs a review cadence and a price, not just patience.
• Confusing a fall from an extreme multiple with cheapness — 53x down to 40x is still not a discount.

4. Look for moats made of certification, where the customer also saves money

1. Identify industries where a third party must approve a product before it can be sold, and where approval takes years and money rather than cleverness.
2. Find the companies that have accumulated many approvals. Each one is a small permanent monopoly, and the stock of them compounds.
3. Confirm the customer is also better off — an approval-based moat that raises the customer's cost invites regulatory or commercial attack; one that lowers it is stable.
4. Check that demand is tied to usage rather than to new-unit sales, so revenue tracks the installed base's activity instead of the capital cycle.
5. Verify the breadth of the catalogue, since a single approved part is a niche and twenty thousand is an infrastructure position.

• Contractual end-runs around the regulation — the risk section names exactly this: "airlines may be restricted from using PMA parts under certain contracts."
• Incumbents defending the aftermarket with bundled service agreements that make the cheaper part irrelevant.

5. Judge a serial acquirer's ratios against its model, and know which thresholds it will always fail

1. Before applying your standard thresholds, ask which of them the business model mechanically violates.
2. For an acquisitive company, goodwill as a share of assets will be high and reported ROIC will be depressed, because every purchase price sits in the denominator.
3. Substitute measures that strip the acquisition accounting — return on tangible assets, return on capital employed, FCF versus net income — and judge on those.
4. Then re-impose the thresholds that still bind: interest coverage, net debt versus free cash flow, and stock-based compensation.
5. Record the failures rather than adjusting them away, so the scorecard remains comparable across business models.

• Goodwill impairments, which convert a tolerated ratio into a realised loss.
• An acquirer whose organic growth is invisible because every reported number is a blend; ask for the same-store figure.

6. Price key-person and single-event risk as its own score

1. Give risk a dedicated line on the scorecard rather than treating it as a discount to the other lines.
2. Name the specific dependency — a family, a founder, a regulator, one customer — and ask what happens the day it is removed.
3. Add the tail event the business model makes possible but rare: for a parts maker, a failure in service; for a lender, a credit event; for a data business, a breach.
4. Score it honestly even when the score is embarrassing next to the rest, and let it constrain position size or the conviction tier rather than the buy/sell decision alone.

• Succession that is announced but untested; the risk score should not fall until the handover has actually happened.
• Reputational tails that are uninsurable — a 5/10 here is not a number you can hedge.

7. Take in outside research, but keep the verdict in-house

1. Accept long-form work from other analysts as raw material — it saves the fact-gathering, which is the expensive part.
2. Re-run your own scorecard and your own valuation on it before forming a view, using the same headings you use for everything else.
3. Publish the summary, the onepager and the score alongside the source document, so a reader can see what was adopted and what was assessed independently.
4. State the decision in your own words, and make it possible for the decision to contradict the contributed case.

• Adopting the contributor's conclusion along with their facts; the length of a document is not evidence for its verdict.
• Undisclosed positions held by the contributing analyst — worth asking before leaning on the work.