Start with the filing, not the headline
Each modeled company has a recorded holding fair value and, where available, a company valuation used as the filing baseline. If an eligible newer company valuation exists, the model multiplies the holding fair value by the ratio of the selected company valuation to the baseline company valuation. If no usable valuation or baseline exists, it keeps the holding fair value unchanged. This is a proportional proxy, not a security-by-security appraisal.
A transparent numerical example
Take an illustrative $40 million holding at the filing date, a $10 billion baseline company valuation and an eligible $12 billion company valuation. The ratio is 1.2 and the modeled holding becomes $48 million. That $8 million change is an input to the fund estimate. It is not an assertion that the fund sold the position for $48 million, that the same price applies to every share class, or that fund shareholders received an $8 million distribution.
Aggregate and divide
The implementation adds the modeled holding values and configured cash, then divides by the configured shares outstanding. It does not independently estimate every liability, fee accrual, dilution event or new transaction. Official NAV is normally a net-assets calculation; this proxy does not have a separate comprehensive liabilities line. A mismatch between the configured baseline and a newer reported NAV can therefore persist. The interface labels the result Estimated NAV and displays the reported reference separately.
Selection is not simply newest first
The existing selector ranks eligible records by audit trust and source class, then by event date within that ranking. A newer headline can lose to an older, more trusted record. Subject mismatches, combined-entity values and offering amounts incorrectly treated as company valuations are excluded by deterministic checks. Large unsupported jumps and old records without corroboration can also be blocked. The source page exposes retained and excluded records rather than silently deleting inconvenient evidence.
Two meanings of confidence
The estimate’s high, medium and low labels are based on the average age of selected company valuation events: below 30 days, below 90 days, and otherwise low. These labels are not probabilities of accuracy. The automated reviewer’s approval threshold is a separate gate for proposed source records. A strong primary source can be old; a recent article can be weak. The model cannot infer missing share-class rights, unknown holdings or future sale proceeds from a confidence label.
Sources and further reading
Educational research, not personalized investment advice. Examples are hypothetical; fund-specific observations carry their own source dates.