[aerohead]

We'd need:
* year model
* test conditions
* pre-test warmup
* testing methodology, instrumentation, data capture
* no 'hypermiling'
* O.E.M. mpg baseline only from that methodology
* modified mpg of modified Prius, captured during identical methodology, with results normalized to standard SAE air density
* large enough 'sampling' to be statistically significant
--------------------------------------------------------------------------------------
The modified shape can be the only variable allowed. Nothing else about the Prius can be changed unless it's independent effects can be parsed out and quantified. Not even tire pressure. Identical weight.
Technically, all testing should take place on the exact same course. Round trip. Or the sample of sufficient magnitude that it represents a sample as large as that used to establish the OEM baseline. The larger the database, the less background noise during comparisons.
Meteorological data must be accounted for.
Road conditions.