The 7-Filter Fund-Selection System

You met the seven filters in the Starter Kit as a shortcut. Now, with Module 2’s mechanics behind you, they become a system you actually understand: each filter encodes one lesson you have already done, which means you can run ANY fund through them, forever, instead of taking my shortlist on faith. This lesson shows the system annotated, the order to run it in, what it deliberately ignores, two full worked examples with October 2026 data, and the mistakes people make when they try to be cleverer than the list.

The system, annotated

Three of the thresholds are judgement calls, so here is the reasoning behind each. 0.25% is roughly double what the large all-world funds now charge (0.07–0.15% in October 2026), so it admits every serious core candidate while rejecting the 0.40–0.80% thematic and “smart” products. €500m is the size above which closures are rare and market makers quote tight spreads; most core candidates are ten to three hundred times bigger. Physical is a default, not a religion: Lesson 2.5 showed where a swap-based S&P 500 fund is a legitimate exception, and Lesson 3.3 names one.

Order matters: cheapest checks first

The filters run cheapest-first. Filters 1–4 are binary reads off a fund page: the word UCITS, the index name, “Acc”, an ISIN beginning IE. Seconds each, and together they eliminate the great majority of the thousands of UCITS ETFs on sale and almost every bank fund. Filters 5 and 6 are two numbers from the factsheet. Only filter 7 and the tracking-difference check need actual minutes, because you have to find the performance table and compare fund against index over one, three and five years.

Applied in order, any fund takes under ten minutes, and the time goes where it matters: on the few funds that survive the first six gates. If you ever find yourself spending an hour on a fund, you have skipped a gate. Go back and run them in sequence.

What the filters deliberately ignore

• Past performance rankings. For index funds tracking the SAME index, performance differences ARE tracking difference (covered by filter 5). For funds tracking different indexes, past ranking is mostly noise (1.1).
• Star ratings and “top fund” lists. Backward-looking popularity contests. Filters 5 and 6 already capture everything durable about them.
• Provider brand wars. Vanguard versus iShares versus Amundi versus SPDR at comparable scale is bikeshedding. The filters treat providers identically and let the numbers speak.
• Perfect optimisation. Two funds passing all seven filters differ by basis points. Passing is good enough; the expensive mistake is delay, not the second-best passing fund (1.3).
• Securities lending yes/no. Lesson 2.6 covered it. It shows up in tracking difference anyway, which is why TD is the final judge rather than a separate lending filter.

Worked example: two funds, seven gates, ten minutes

Fund A: Vanguard FTSE All-World UCITS ETF (USD) Accumulating, VWCE / VWRP, ISIN IE00BK5BQT80. Data from the Vanguard UK product page, portfolio as at 31 August 2026.

1. UCITS — yes, it is in the name. Pass.
2. Broad global index — FTSE All-World, 3,784 holdings across developed and emerging markets. Pass.
3. Accumulating — yes (“Accumulating” share class). Pass.
4. Irish domicile — ISIN starts IE, domicile Ireland. Pass.
5. TER — 0.14% ongoing charges figure (cut from 0.22%; always read the current page). Pass, with the TD check to follow.
6. Size — US$87.29 billion across the fund, US$60.19 billion in this share class. Pass by a factor of more than a hundred.
7. Replication — physical. Pass.

Seven of seven. Elapsed time: about four minutes, most of it finding the right page.

Fund B: iShares Global Clean Energy UCITS ETF, INRG, ISIN IE00B1XNHC34. Data from the iShares UK product page, 6 October 2026. This is a popular, well-run fund from a top issuer — exactly the kind of product people assume must pass.

1. UCITS — yes. Pass.
2. Broad global index — S&P Global Clean Energy Transition Index, 106 holdings in one sector. Fail.
3. Accumulating — no, it distributes semi-annually. Fail for an accumulation-phase investor.
4. Irish domicile — yes. Pass.
5. TER — 0.65%. Fail (2.6 times the ceiling).
6. Size — US$3.49 billion. Pass.
7. Replication — physical. Pass.

Three failures, and you did not need to look at a single return figure to find them. If you do look, the calendar returns on the same page explain why filter 2 exists: +140.2% in 2020, then −24.1%, −5.6%, −20.5% and −26.1% in the four following years, then +46.0% in 2025. A diversified world fund lost 17.7% in 2022 and had no other losing year in that run (FTSE All-World, USD, FTSE Russell factsheet 30 September 2026). The system is not saying clean energy is a bad idea. It is saying you cannot build a core on a product whose annual swing is routinely 25 points in either direction.

The tracking-difference check, done properly

Filter 5 has two halves. The TER is the sticker; the tracking difference is the receipt — what you actually kept relative to the index after costs, withholding tax, lending income and the manager’s skill at rebalancing. You read it from the factsheet’s performance table: fund return minus index return, ideally over three and five years, because single years are noisy.

Here is what that looks like in October 2026, from the issuers’ own factsheets dated 31 August 2026. The State Street SPDR MSCI All Country World ETF (SPYY, TER 0.12%) returned 22.51% net over one year against 22.35% for its index: a difference of +0.16 percentage points, and +0.10 a year over three years. The Invesco FTSE All-World (FWRA, ongoing charge 0.15%) made 22.32% against 22.29%: +0.03. Amundi Prime All Country World (WEBN, 0.07%) made 22.46% against 22.40%: +0.05. All three BEAT their index after fees. That is possible because the index assumes the maximum withholding tax on dividends, while an Irish fund actually pays less (Lesson 2.2) and earns a little from lending (2.6). The lesson: a 0.15% fund can out-track a 0.07% fund in a given year. The sticker gets you to the shortlist; the receipt picks the winner, and even then the differences are hundredths of a percent.

Source: issuer factsheets, performance to 31 August 2026, net of fees, in USD. One year is shown because WEBN only launched in June 2024; prefer three- and five-year figures when a fund has them.

Where the thresholds bite: the cost of ignoring filter 5

To see what the 0.25% ceiling protects, take the Lesson 1.4 investor: €300 a month for 30 years, 6% a year before costs, costs deducted from the return (illustrative, not a forecast). At 0.07% the pot ends near €297,000; at the 0.25% ceiling, near €287,000; at a typical bank-sold active fund’s 1.5%, near €228,000. The ceiling costs you about €10,000 against the cheapest fund. The bank fund costs you about €70,000. That asymmetry is the whole reason the filter is a ceiling rather than a race to the bottom: the first 1.25 percentage points of fee reduction matter enormously, the last 0.18 matter a little.

If you want to run your own numbers, the free ETF fee calculator on the site does exactly this comparison for three TERs at once.

Common mistakes

• Treating 0.25% as a target instead of a ceiling. People find a 0.20% fund, see it passes, and stop. Fix: run the TD check; among passing funds, prefer the one with the better three-year receipt, then the lower sticker.
• Running filter 7 before filters 1–4. You spend twenty minutes reading a replication note on a fund that is Luxembourg-domiciled and distributing. Fix: order is the system. Binary gates first.
• Confusing “global” with “broad”. A global clean-energy fund, a global robotics fund and a global dividend fund are all global; none is broad. Fix: filter 2 means the whole investable market, cap-weighted: thousands of holdings, every sector.
• Letting fund age override size. A two-year-old fund with US$8 billion (WEBN) is not a risky fund; a fifteen-year-old fund with €40 million is. Fix: size and tracking difference are the data; age is a comfort feeling.
• Applying the equity domicile rule to bonds. Filter 4 is about US dividend withholding. Bond interest is generally not withheld the same way, so a Luxembourg money-market ETF is not a failure (Lesson 3.5 covers this).
• Re-running the filters every month. They are a purchase gate and an annual audit, not a hobby. Fix: once at purchase, once a year in the 30-minute checklist (6.6).

Checklist: one card, seven lines

1. UCITS in the name — yes/no.
2. Index covers the whole world market, cap-weighted, thousands of holdings — yes/no.
3. Share class is accumulating — yes/no (distributing only if Lesson 2.3 said so for you).
4. ISIN starts with IE — yes/no (equity only).
5. TER at or below 0.25% — write the number. Then: fund minus index over 3 years — write the number.
6. Fund size at or above €500m — write the number and the as-of date.
7. Replication physical — yes/no; if synthetic, write the one-sentence reason it is allowed.

Seven yeses and two written numbers: the fund is eligible. Anything else: it is not, however good the story. You will use this card three ways — choosing your core fund (next lesson), auditing anything anyone ever pitches you (“it fails my filter 5, thanks”), and re-checking your own holdings every year.

Sources checked (October 2026): Vanguard UK, FTSE All-World UCITS ETF (USD) Accumulating product page (portfolio as at 31 Aug 2026), iShares UK, Global Clean Energy UCITS ETF product page (6 Oct 2026), iShares UK, Automation & Robotics UCITS ETF product page (6 Oct 2026), State Street SPDR MSCI All Country World UCITS ETF factsheet (31 Aug 2026), Invesco FTSE All-World UCITS ETF Acc factsheet (31 Aug 2026), Amundi Prime All Country World UCITS ETF Acc factsheet (31 Aug 2026), FTSE Russell, FTSE All-World Index factsheet (30 Sep 2026). Figures are illustrative and change; verify before acting.