Actionable insights — ETF Portfolio Update August 2026

1. Turn a soft factor into the smallest set of computable inputs, and defend each one with its own evidence

1. Write down what you mean by the factor in words first (here: reinvests at high rates, survives, earns real cash).
2. Assign each clause exactly one measurable input — return on equity, financial leverage, accruals ratio.
3. Produce a separate piece of evidence that the input actually predicts returns, and read the shape of that evidence, not just its direction.
4. Combine into a single score, then cut to a fixed count rather than a threshold, so the screen always returns a workable list.
5. Decide the weighting rule separately from the selection rule — here score × market cap.

• Asymmetric inputs used symmetrically. If low debt does not help, scoring companies upward for it adds noise; it belongs as a disqualifier, not a score component.
• Three inputs is few enough to be robust and few enough to be gamed. Ask what a management team optimising for exactly ROE, leverage and accruals would look like — buybacks funded by debt raise ROE while worsening leverage, and the two partly cancel.

2. Start portfolio design from the return distribution, not from the average

1. Look up the median outcome alongside the mean for whatever population you are about to buy.
2. If the two diverge violently, treat the average as a description of a few names rather than of the population.
3. Ask what a cap-weighted index actually buys under that distribution: everything, including the majority that loses money.
4. Decide whether your response is to concentrate (own fewer, better) or to accept the average (own it all) — and be explicit that this is the choice being made.

• The unstated counter-argument: skewness is also the reason indexing works. If the winners are unpredictable, owning everything guarantees you own them; a filter that drops 400 names can drop the one that mattered.
• A screen justified by skewness that is itself cap-weighted, which reintroduces the concentration it set out to fix.

3. Run the same policy twice, in two wrappers, and use the pair as a natural experiment

1. Where a regulatory split forces two implementations of one strategy, map every position to its counterpart explicitly.
2. Chart the pairs side by side on the same axis, so a difference in outcome is visible rather than buried in two separate reports.
3. When a pair diverges materially, diagnose it: is it currency, index construction, geographic scope, or fee?
4. Write the diagnosis down. A gap left unexplained becomes an unexamined bet.

• "World" versions of a US strategy quietly becoming a currency and geography bet — most of these gaps are the dollar and the US weighting, not manager skill.
• An expectation of convergence stated as a reason not to investigate. Convergence over decades is compatible with a permanently worse implementation.

4. Read a rules-based fund by its holdings, not by its name

1. Pull the top ten and the sector split before accepting the label on the tin.
2. Ask which businesses the screen's inputs would mechanically favour right now, given where each industry sits in its own cycle.
3. Flag any holding whose high score depends on peak-cycle earnings — a trailing-ROE screen buys cyclicals at their best moment by construction.
4. Cross-check the fund's largest positions against views you already hold on those same names elsewhere in your book.
5. Size the position for the holdings you actually get, not for the factor you meant to buy.

• Turnover in the fund: how quickly does the screen drop a name whose ROE collapses? A once-a-year rebalance means owning the downswing.
• Overlap with what you already own directly — Visa and Mastercard are 9.75% of this fund, and Visa is separately a portfolio position.

5. Audit each monthly letter for what has silently disappeared

1. Keep a running list of every position and every "of the month" pick as it is published.
2. On each new issue, diff the current holdings chart against your list before reading the prose.
3. Any name that vanished without a sell note is the highest-value question in the issue.
4. Ask for the exit price and reason; absent that, treat the strategy's published record as incomplete.

• The benign explanations — a chart that only shows the largest positions, a renamed fund (the same issue labels one holding "Tema Durable Quality" in one chart and "Tema Monopolies and Oligopolies" in another) — which are just as important to confirm as a quiet sale.
• Performance figures that would change if a dropped position were included.

6. Compare the weighting order against the performance order every month

1. List the positions twice: once by weight, once by return.
2. Where the two orders disagree sharply, ask whether the weighting expresses conviction, inertia, or a deliberate contrarian tilt.
3. Check whether new money follows conviction or follows performance — adding to a winner and adding to a laggard are different policies and should be conscious.
4. Re-read the sleeves the strategy claims to run against the weights actually deployed.

• A "quality" letter whose biggest position is a size factor — the stated sleeves and the actual weights should be reconciled explicitly.
• ETF adds executed at the open with no limit, while the equity book publishes named limit prices. Two standards for the same money.