Subset-weighting: why missing data drops out
Why Veridion drops missing factors instead of filling them with neutral values.
The missing-data problem
When a factor has no usable data for a ticker, Veridion drops the factor from the composite rather than filling it with a neutral placeholder. This choice matters.
A ticker can lack analyst coverage, have no recent insider activity, or have too little trading history for a clean momentum read. Treating that missing field as neutral would create false precision.
The two choices
One approach is neutral-fill. The missing field receives a neutral value and the composite proceeds as though every factor existed.
The Veridion approach is subset-weighting. The missing factor is dropped, and the remaining factors are reweighted across the available subset.
Neutral-fill produces a smoother number. Subset-weighting produces a more honest number.
Why Veridion uses subset-weighting
Every neutral-filled score says, quietly, that the missing factor had information. It did not.
Subset-weighting keeps the math aligned with the data. If a factor has no usable input, the composite is computed from what is known. The confidence badge then tells the reader how complete the coverage was: high, medium, or low.
The headline number never imputes a neutral fallback where data is missing. Sparse coverage produces a less-confident read, not a padded headline.
A worked example
Two tickers can both show a score near the same level.
Ticker A has all six factors present. Valuation, hype, earnings, sentiment, analyst rating, and momentum all contribute. Confidence is high.
Ticker B has two factors present. Earnings and momentum contribute; the other four factors have no usable data. Confidence is low.
The numeric score may look similar. The confidence badge tells a different story. Treating the two reads as identical would hide the coverage difference.
Where the rule applies
Subset honesty applies across Veridion surfaces.
- The Veridion Score drops factors without data
- Forward-return evidence drops rows without complete windows
- Portfolio aggregation drops holdings without a current Score and discloses the dropped set
- Risk metrics are omitted when the underlying bar history is too sparse
The rule is consistent: missing data is dropped, never zero-filled.
Closing note
When you see a score, read the band and the confidence together. Both matter. The math treats missing data as missing data, not as a quiet default.
Not financial advice. Just receipts.