How we measure a fund against its benchmark

01 · Active share

Difference is not skill

Active share is the portion of a fund portfolio that differs from a benchmarkportfolio. A result of 70% means 70% of the two portfolios does not overlap at the measured weights on that as-of date.

Active share can’t tell you whether the active bets were good—only that they exist. It does not measure future returns, manager skill, risk, or whether the fee was worth paying.

02 · Benchmark rule

The fund gets the most charitable comparison

We calculate active share against 8 index portfolios and publish the lowest result as the best-fit headline. Each index portfolio uses weights from the ETF’s filed N-PORT holdings nearest the fund’s fiscal-quarter date.

We also calculate active share against the benchmark named in the fund’s prospectus and show that number beside the best-fit result. When the declared benchmark and best fit disagree, the registry marks the row “benchmark mismatch.”

Best-fit can understate activeness for funds that genuinely straddle styles—the mismatch flag exists for exactly that case. The lowest result is conservative, but it is not proof that the best-fit index describes the manager’s mandate.

03 · Fund overlap

The shared weight in two filed portfolios

Fund overlap adds the smaller of the two portfolio weights for each shared issuer. The comparison uses long common equity, collapses multiple share classes of one issuer under the same CUSIP-6 identity, and remains keyed to the published run shown with the result.

Class tickers resolve through the permanent fund registry. For each resolved fund, the comparison uses its latest holdings-bearing source run at or before the pinned run. The quiet provenance line gives the pinned run; its details name each fund’s holdings source run and filing date.

A percentage appears only when both funds have at least 70% CUSIP-6 coverage. Below that line, the result says the fund is under review and shows no number. A ticker outside the current tracked set says “not yet tracked.”

The result lists up to 10 shared positions and 5 positions found only in each selected fund. A visual needs at least 3 shared positions; below that threshold, the page uses a compact sentence and a small filed-positions table.

The two filing dates can differ, and either fund may have traded since its quarter-end snapshot. Overlap measures shared filed weight. It does not measure quality, future returns, risk, or whether either fee was worth paying.

03A · Portfolio X-ray

Amounts weight the available fund facts

The X-ray resolves class tickers and permanent aliases to the current registry run. An amount outside that run is labeled not yet tracked and excluded. A missing expense ratio, licensed return window, or active-share quotient is also excluded from that figure and reported in dollars; absence never enters an average as zero.

The blended expense ratio is the sum of each included amount multiplied by its filed class expense ratio, divided by the included amount. Convention fee drag adds each holding’s 10-, 20-, or 30-year difference from the computed index-control fee under the 5.0% annual-return convention. Real-window fee drag scales and adds only complete licensed trailing windows, retaining the “Computed on licensed market-return history” label.

Weighted fee per active unit uses amount weights only where the registry publishes that quotient. A fund below 5.0% best-fit active share remains suppressed under methodology 1.2. Pairwise cells come from the same run-keyed overlap computation described above and retain its identifier gate.

Tickers are sent in pairs to the overlap service. Dollar amounts remain in page memory, are not placed in those requests, and are not written to URLs, cookies, browser storage, accounts, or a server. The share block copies plain text and creates no saved state.

A weighted total can hide a missing fund behind the available portion. Read the included and excluded dollar amounts beside each figure. These source-based illustrations are not forecasts and do not determine whether a fund fits an investor.

04 · Portfolio filters

What enters the comparison changes the answer

The comparison includes long common equity. It excludes preferred shares and derivative overlays. The Phase 1 evidence review from July 25, 2026 recorded that those exclusions shifted the comparison base by up to roughly 5.5% in its validation sample. That observed limit is not a forecast for the registry; each fund’s excluded weight must stand on its own filing.

We collapse dual classes to the issuer level with a curated map, including Berkshire Hathaway and Alphabet share classes. A manager does not earn active share merely by holding a different class of the same issuer. Funds-of-funds stay outside the registry because comparing their wrapper holdings with direct equity weights would mix two different portfolio levels.

These filters make common-equity comparisons more consistent, but they can hide risk or exposure created by excluded preferreds and derivatives. Read active share as a long-common-equity measure, not a complete account of the fund.

05 · Filing delay

The fund has traded since the snapshot

The Phase 1 evidence review from July 25, 2026 records that fund holdings filings can arrive up to roughly 60 days after quarter end. Each registry figure therefore shows the as-of date from the underlying filing, not the date a reader opens the page.

The portfolio may have changed before the figure appears. Stale is acceptable; undisclosed stale never is.

06 · Fee metrics

Filed fees, divided by measured difference

Net expense ratios come from the risk/return summary in the fund’s filed prospectus data. Marketing pages are not a fee source.

Fee per active unit divides the filed net expense ratio by best-fit active share. A 1.00% expense ratio divided by 50% active share equals 0.02 percentage points of annual fee for each percentage point of measured active share. It is a comparison unit, not a second fee charged to an investor.

When best-fit active share is below 5.0%, the fee-per-active-unit quotient stops being useful. The verdict instead compares the fund’s annual dollars per $100,000 with the cheapest published fund class below 5.0% active share against the same best-fit index. A more expensive tracker names the lowest-cost fund, gives both annual dollar fees, and gives the 20-year difference under the fee-drag assumptions below. When the headline class is itself the lowest-cost tracker, its verdict states that status and no comparison row is shown against itself.

In an index-tracker verdict, “essentially the same portfolio” has a narrow computed meaning: the fund and named control each have less than 5.0% active share against the same best-fit proxy. This is not a direct pairwise holdings comparison.

A lower ratio does not make a fund better, and a higher ratio does not prove poor management. The measure cannot value the active decisions. Dollar fee-drag illustrations use the version 1.1 assumptions below and do not measure whether the fund’s decisions justified its fee.

07 · How to read a fund page

Start with the verdict, then check the inputs

A verdict uses the page’s best-fit active share and the headline share class’s filed expense ratio. Below 5.0% active share, it leads with annual dollars per $100,000. From 5.0% through 59.9%, it gives a closet-index warning only when the filed fee exceeds the cheapest same-index fund, stating how much of the portfolio matches the index and both annual dollar fees. When the fund costs no more than that fund, the same plain-dollar comparison appears without the warning. At 60.0% or more, the verdict reports active share, the filed fee, and the limit that holdings cannot show whether the fund earned that fee. A fund whose name identifies it as an index fund stays in the index-comparison wording even when it differs substantially from the closest broad-market index in this registry.

The line is arithmetic, not a rating. Read the portfolio and expense as-of dates next, then the largest active differences and limitations. A verdict is unavailable when the identifier gate, active-share input, expense filing, or same-benchmark control needed for that arithmetic is unavailable.

“Genuinely active” means 60.0% or more best-fit active share under this rule. It does not mean the manager was skillful or that the fee produced a better result.

08 · Record policy

Published marks are append-only

A published as-of mark is never rewritten. A later filing creates a later mark. If we confirm an error, the correction appends to the record and this changelog identifies what changed; it does not erase the original observation.

Each verification URL resolves to the record for that as-of date. A current registry page may show a newer filing, but its older verification URL remains a citation for the older mark.

Append-only history preserves mistakes as well as correct results. The correction record is part of the evidence readers need to judge either one.

09 · Fee-drag illustrations

The difference between two fixed fee rates

Each illustration starts with $100,000, applies a 5% annual return, compounds annually, and deducts the filed expense ratio once per year as a reduction of return. The 5% rate is the comparison convention required by the SEC’s Form N-1A fee-example instructions, not our estimate of future performance.

The calculator reports ending value at the fund class’s filed net expense ratio, ending value at the baseline, and the difference. The baseline is the lowest filed expense ratio among the registry’s versioned index-fund controls and is recomputed when the registry snapshot changes. The current baseline is0.01%. Static tables use 10, 20, and 30 years; the calculator allows 1–50 years.

The current 0.01% baseline is supported by VSTSX, filed expense period 2025-12-31; VFFSX, filed expense period 2025-12-31; SSEYX, filed expense period 2025-12-31.

The arithmetic holds returns and expense ratios fixed. Actual returns and fees change, taxes and cash flows are omitted, and the result is not a prediction or personalized advice. A smaller illustrated difference does not establish that a fund is suitable or that its portfolio decisions were successful.

10 · Licensed return windows

Actual history, kept separate from the SEC convention

The registry has two provenance classes. Active share, filed expense ratios, and the fixed-return fee illustration are Computed from SEC filings. Real trailing-window fee drag and tracking difference are Computed on licensed market-return history. The labels stay with the figures so a historical result is never presented as an SEC-convention result.

Real-window fee drag uses the best-fit proxy’s adjusted-close total-return history over complete trailing 10-, 20-, and 30-year windows. It applies the class’s filed expense ratio and the same lowest-filed-index-control baseline used by the convention illustration. Each result shows its actual start and end date. A shorter series remains null and says when proxy history begins; we do not pad or extrapolate it.

For index-control mutual-fund classes with provider NAV history, tracking difference is the fund NAV’s annualized return minus the matched proxy’s annualized return over exact shared dates. Only the derived annualized percentage is published.

The underlying market-return and NAV series are licensed and remain inside the backend. Raw prices and returns cannot be republished, exposed through a public API, or offered as a download. That is why readers cannot download those raw series here. The site publishes only derived percentages and dollar outcomes.

11 · Visual reading rules

The map describes; the fund rules decide

Visual explainer rules version si-visual-explainers@1.0.0

The Closet Index Map places best-fit active share on the horizontal axis and the headline share class’s filed net expense ratio on the vertical axis. The 60.0% active-share line is also where the existing verdict changes wording; the map does not assign that wording to a region. The horizontal divider is the median filed fee among the funds with both plotted inputs. Its four regions state those two facts only. A region is not a verdict and does not trigger a warning.

The readable cross-section selects up to three names in each region after sorting by active share, fee, and permanent fund identity. It exists to keep direct labels readable. The complete table, not the cross-section, shows prevalence.

A Vs-Universe percentile equals 100 times the number of lower current values plus half the number of equal values, divided by the number of available values. The fund itself is included. Active share uses published funds with active share; fee uses published funds with a headline-class fee. Each result names its comparison count and source date.

Rule 1 · Declared-benchmark mismatch

The warning appears only when the fund’s filed benchmark resolves to one approved comparison and the mechanical best fit resolves to another. A match clears the check. An unresolved or unfiled benchmark leaves it unavailable.

Rule 2 · Closet-index warning

This is the existing fund-verdict rule, not map-region membership. From 5.0% through 59.9% best-fit active share, it warns only when the filed headline fee exceeds the cheapest published fund below 5.0% active share against the same best-fit index. The existing named-index treatment remains part of that verdict computation. Tracker wording applies below 5.0%; 60.0% or more uses the genuinely-active wording unless the named-index rule applies.

Rule 3 · Licensed tracking-difference check

For the headline class, this check takes the longest complete licensed window; ties use the latest end date and then latest start date. Tracking difference is signed annualized fund return minus annualized index return. The warning appears only when that value is less than the negative filed fee—meaning the annualized lag is greater than the fee. Equality clears the check. Missing fee or complete history leaves it unavailable. The result shows its exact window and remains Computed on licensed market-return history, display-only.

A clear check is not a recommendation. Active share does not measure skill, and tracking history does not predict a future gap.

12 · Changelog

Method changes remain visible

A future methodology change appends here with its version and date. It does not silently alter the meaning of a previously published mark.