Actionable insights — Portfolio Update: Four-stock Edition

1. For a cyclical, value on Price/Sales — a peak-cycle P/E is cheapest exactly when the earnings are least repeatable

1. Classify the business first. If revenue and margin are driven by a commodity price or a capacity cycle the company does not control — energy, mining, shipping, chemicals, semis-equipment — treat it as cyclical and stop reading its P/E as a valuation.
2. Understand the failure mode before you replace the ratio: at a cycle peak, the denominator of P/E is a margin that will not persist. The ratio therefore prints its lowest, most tempting number at the worst possible moment to buy, and its highest number at the trough, when the earnings are temporarily crushed.
3. Substitute Price/Sales. Sales fluctuate with volume and price but not with operating leverage, so the series is far smoother and much harder to flatter with accounting.
4. Compare the P/S to the company's own history, not to the market or to a different sector — a five-year band of its own readings, so "cheap" means cheap relative to how this business has been priced through its own cycle.
5. Sanity-check the two ratios against each other at a known cycle peak. If P/E collapsed far more than P/S did in that window, you have confirmed the distortion is present in this name and the P/S reading is the one to act on.

• A cyclical screening as a single-digit P/E after a commodity spike — that is the distortion, not an opportunity. Conversely, watch for the mirror-image error: refusing a cyclical because its P/E is enormous at the trough, when earnings are temporarily near zero. Also check that sales themselves are not being inflated by an acquisition, which breaks the P/S comparison to history just as badly.

2. Trim a winner only when one of three things has deteriorated — the size of the gain is not one of them

1. Start from the default. A position that has run hard is a candidate for trimming — treat "we often suggest trimming positions that have been extremely rewarding" as the standing rule, so the exception has to be argued.
2. Then run the three-part test on the current state, ignoring the entry price entirely: is the chart pattern still constructive, is the valuation still undemanding on the right metric, is the fundamental story still intact?
3. All three still encouraging → hold the full position and let it run. The gain itself carries no information about what happens next.
4. One or more deteriorated → trim, and say which one broke. Distinguish the failure modes: a broken fundamental story argues for exiting; a stretched chart or multiple with the story intact argues only for reducing size.
5. Never let optionality do work in the decision. Name any takeover or catalyst possibility, then explicitly discount it so no part of the position rests on it.

• The two symmetrical errors this test is designed to prevent: trimming a compounder purely because it is up (nothing failed), and holding a parabolic name because "the story is still great" (the story was never the thing that broke). Watch for the tell that the test is being skipped — a trim justified by the gain alone, or a hold justified only by the thesis.

3. Measure "extended" against the sector index, not against your eye

1. Identify the index or ETF that best represents the driver of the position — the sector the name belongs to and gets its beta from.
2. Express both at the same reference point: how far is the single name from its own prior high, and how far is the sector from its prior high?
3. Read the gap. A stock back at its peak while its sector is still meaningfully below its own is outrunning the thing that drives it — that is measurable stretch rather than an aesthetic judgement about a chart.
4. Ask why the gap exists before acting on it. If there is a durable structural reason (unique index membership, a franchise asset, a balance-sheet advantage) the premium may be legitimate — but a legitimate premium that keeps widening is still a sizing question.
5. Cross-check with a second, independent read — the shape of the recent advance and the valuation multiple — so the trim rests on more than one observation.

• The signal reversing: the single name stalling while the sector index makes new highs is the pattern that says the premium is being handed back, and it usually arrives before the fundamentals say anything. Watch too for picking the wrong benchmark — a senior-miner ETF is the right comparator for a senior miner, not a bullion price or a junior index.

4. Pre-commit to the price that confirms a breakout — and hold the call open until it prints

1. Identify the resistance level from its history: the price at which prior advances repeatedly stopped, and the episode that created it. The older and more tested the ceiling, the more supply has been absorbed when it finally breaks.
2. Distinguish the near-term ceiling from the structural one. Clearing a recent range is a smaller event than clearing a multi-year one, and they can happen months apart.
3. When the break happens, do not declare it confirmed. State plainly that it is "not yet decisive" and name the specific price that would confirm it.
4. Prefer a confirmation defined as trading through and holding above the old ceiling zone, since the level that capped the stock becomes support only once it has been defended from above.
5. Keep the valuation test running in parallel. A breakout with the multiple still undemanding is a position to hold; a breakout that has already consumed the valuation case is a trade, not an investment.

• The failed break — a push above a multi-year ceiling that cannot hold it on a retest is a stronger sell signal than never breaking out at all, because it proves the supply is still there. Watch also for a breakout that arrives only after the valuation case has gone; at that point the chart is the entire thesis.

5. Re-underwrite a loser by asking whether the evidence improved, not whether the price did

1. State the loss and the entry without softening. The review is worthless if it starts by managing the reader's (or your own) feelings about the drawdown.
2. Write down the original thesis in one sentence, as specifically as it was made — not a vaguer version that is easier to keep defending.
3. Test that exact sentence against everything that has been published since. The question is whether the evidence for it is stronger or weaker today, which is independent of the share price.
4. Decompose a bad headline number into its causes. Spending that depresses reported profit because it is building the asset the thesis rests on is a cost of the thesis working, not evidence against it — but you must be able to point at the operating metric that shows the asset is actually being built.
5. Demand a commercial link between the strategic evidence and the revenue line. An adoption statistic is only a thesis-advancing fact if you can name the mechanism by which adoption becomes money.
6. Then price it. Concede the discount is real, ask explicitly whether it is deserved, and make the call on that question alone.

• The failure mode this method invites — an infinitely elastic thesis, where every miss becomes "investment" and every quarter is early. Guard it by requiring the operating metric to keep accelerating (here, cloud growth and download share): the day the spending continues but the adoption metric flattens, the capex explanation is gone and the position is simply wrong.

6. Separate the rating from the holder — publish the suitability condition alongside the call

1. After settling the rating, ask a second and separate question: who can actually hold this position through what it is likely to do?
2. Name the two requirements concretely — the horizon the thesis needs, and the drawdown the holder must be able to sit through — rather than a generic risk warning.
3. Give the people who fail that test an explicit alternative instruction, so "buy" does not read as universal advice.
4. Publish your own base rate somewhere in the same note. A stated hit rate tells the reader how much weight any single call deserves and pre-commits you to counting the misses.
5. Keep the housekeeping visible and separate from the ratings — a corrected cost basis or trade date is bookkeeping, and labelling it as such prevents it being read as a new recommendation.

• A "buy" issued with no suitability condition on a name whose thesis needs years and whose chart moves 20% a quarter — that is a rating written for the analyst's scorecard rather than for a holder. Watch too for a research provider whose published hit rate never appears, or only appears in the winning quarters.