Cases · Report check · finance
A fund factsheet whose top 10 does not reproduce from the fund's own holdings
Self-initiated case study on public documents, not synthetic data: an iShares fund's factsheet and the fund's own holdings files. The issuer's files are not republished here.
Not a finding of an error: the printed weights may rest on another basis or another day the document does not name.
Goal
Take a public fund factsheet as the report and the fund's own published holdings as the data, and find out which printed figures can be recomputed from the holdings and whether they come out the same. The fund: iShares Core MSCI Europe UCITS ETF EUR (Acc), ticker SMEA.
Challenge
The factsheet names two dates: 31 August 2026 for portfolio breakdowns and 3 September for all other data. Only 11 of its 52 figures can come from a holdings file: the ten largest positions' weights and their sum. Returns, fees, net assets and beta cannot. And the 3 September file changed shape: currency forwards came as several lines with the same ticker and name.
Solution
The basis of a weight was agreed before counting: percent of the market value of every line of the file, as the file's own weight column is computed. The 11 figures were recomputed from the holdings file of 31 August (405 lines, 387 of them shares), then recounted from the file of 3 September. Each of the other 41 figures is listed with the reason it cannot be recomputed. On the 3 September file the checks stopped instead of pairing lines they could not tell apart.
Outcome
10 of the 11 do not reproduce, on 31 August or on 3 September. The top 10 add up to 21.02 on the factsheet, 20.98 in the 31 August holdings and 21.13 in the 3 September holdings. The factsheet prints 396 holdings; the file has 387 share lines, and 396 only with its 9 cash lines added.
Source documents
Factsheet (iShares, PDF) · Holdings, 31 Aug 2026 · Holdings, 3 Sep 2026
iShares replaces the factsheet every month, so the link leads to the latest one.
From the case
The case, 1 page (PDF)
Service
Report checks for your own files
The same method on your exports: definitions in writing first, then an independent recount.