Actionable insights — 10 Stocks for the next 20 years

1. Impose an artificial lock to change what you are optimising for

1. Set the exercise up precisely: a fixed number of individual companies, no funds, no additions, no sales, dividends reinvested, for a stated multi-decade period.
2. Notice what the lock removes — the ability to correct a mistake — and let the objective change accordingly, from maximising return to minimising the chance of a permanent loss.
3. Write down every name you would actually be comfortable never revisiting. The list will be much shorter and much duller than your real portfolio.
4. Compare the two lists. The gap between them is the part of your portfolio that depends on you being able to change your mind, and it should be sized as such.
5. Repeat periodically — the exercise is a calibration tool, not a portfolio construction method.

• Smuggling the constraint away by picking holding companies and conglomerates that can trade on your behalf — legitimate (it is why Berkshire ranks first) but it should be a conscious choice, not a loophole.
• A locked list that is really a low-volatility list; durability and low beta are not the same property.

2. Ask one question of every candidate — "why will this still be relevant in 20 years?"

1. Force the answer into two headings only: how the company makes money, and why that mechanism still exists in two decades.
2. Reject answers that describe the present ("growing fast," "great margins") rather than the future's structure.
3. Demand that the durability argument be about something outside the company's control — a permit regime, a legal requirement, a biological fact, a physical network — rather than about management's excellence.
4. Write the answer in three bullets. If it takes more, the case depends on a chain of events rather than a standing condition.

• Durability arguments that are really industry-growth arguments; the question is whether the company still takes its cut, not whether the market is bigger.
• Regulatory moats being re-legislated — the same rule that creates the toll can remove it.

3. Prefer networks that could not be built again today

1. Ask literally: if a competitor were handed the capital tomorrow, could they reproduce this asset? If the blocker is money, it is not a moat. If the blocker is permission, geography or time, it is.
2. Look specifically for assets whose supply is politically constrained and shrinking — permits, licences, land near cities, grid connections.
3. For service networks, measure density rather than size: stops per route, stores per market, the marginal cost of the next customer on an existing round.
4. Check that the density advantage shows up in economics (cost per unit served), not just in a map.

• A network being routed around rather than replicated — the entrant does not rebuild your stores, it removes the need to visit one.
• Density advantages that decay when a large customer in-sources the service.

4. Rank demand by how little choice the customer has

1. Classify the revenue: is the purchase required (by law or by the system), necessary (the alternative is unacceptable), or preferred (the customer likes it)?
2. Weight required and necessary far above preferred for a long-horizon holding — preference is the thing most likely to change over twenty years.
3. Where demand is required, identify the rule that requires it and how stable that rule has been historically.
4. Treat rising regulatory complexity as a tailwind for whoever absorbs it on the customer's behalf, and check that the company is positioned as the absorber.

• Required demand with a regulated price — compulsion without pricing power is a utility, not a toll.
• Necessity that is really habit; the test is what happens in the first bad recession, not in a survey.

5. Over a long horizon, underwrite the inflation clause, not just the growth rate

1. For any holding meant to survive decades, find out whether its prices reset with inflation automatically, by negotiation, or not at all.
2. Prefer contractual escalators over pricing power that must be re-won every year, and note the contract length.
3. Pair the escalator with an asset that cannot be duplicated, so the counterparty has no alternative when the escalator bites.
4. Then ask the cost side: does the business also have costs that inflate faster than its prices (labour-heavy service models are the usual offender)?

• Escalators capped below actual inflation, or linked to an index that diverges from the company's own cost basket.
• Contract renewals clustered in one period — twenty-year safety with a single re-pricing cliff in year twelve.

6. Know when you have run a durability screen and still owe yourself an entry price

1. Recognise that a "own it forever" exercise deliberately suppresses valuation — with no ability to trade, timing has no value, so it is excluded by design.
2. Treat the output as a candidate universe: names that pass the durability test and are therefore worth pricing.
3. Re-impose the entry discipline separately — a multiple versus the company's own history, or a reverse DCF, before any actual purchase.
4. Keep the two verdicts distinct in your notes, so "great business" is never quietly read as "buy."

• Durability lists quietly becoming buy lists when the market falls — the price test still has to be run.
• Paying a permanent premium for permanence; a twenty-year business bought at forty times earnings can still deliver a poor twenty-year return.

7. Read the exclusions — what a durability list leaves out is the finding

1. After building the list, write down the sectors that produced no entries at all.
2. For each absence, state whether it reflects genuine fragility (competitive position under active renegotiation) or simply unfamiliarity.
3. Check the list for hidden concentration — geography, currency, one economy's regulatory regime — that a "diversified" ten names can easily hide.
4. Publish or record the list and invite disagreement; a durability claim is a falsifiable one.

• Home-country bias masquerading as durability — irreplaceable networks exist in other jurisdictions too.
• Excluding an entire sector because it is currently being disrupted; twenty years is long enough for the disruption to resolve either way.