Fuel burn is very much a metric that is used for selecting aircraft and is part of a weighted scale when it comes to selecting the appropriate size and type.
Airlines like most large businesses use NPV analysis when investigating new investments in equipment. Fuel is only a part of the factors, every airline network is different. Finance (lease terms on the airframe, engines, apu) and maintenance can very well be the determining factor. Flleet commonality and maintenance planning are large costs to an airline over 10-20 years they are plan to operate them.
If an airline already operates A330s adding A330-900s will have different level of investment for training and maintenance than adding a new type like the 787. None of that is related to fuel.
NPV only is applicable to capital financing and not related to operational costing. Predictive fuel models are used all the time when making procurement decisions and you don't have to go into the details to make good models.
When we do fleet planning analysis we do NPV but also apply a methodology of fuel forecasting fuel costs on the proposed fleet. This is also in conjunction with crewing models.
For Fleet Replacement
Based on historical
1) On a flight by flight basis determine obtain fuel from ACARS uplift and fuel burn data
2) Normalize fuel actual burn data to ASMs by flight
3) Build historical fuel burn data distributions per ASM by fleet type
4) Apply ARIMA (autoregressive integrated moving average) analysis based on previous data to future schedule.
-Determine fleet burn numbers (by time series, total monthly amount)
5) Error check with existing fuel burn numbers by fleet type on previous flown schedule.
-total amount of fuel uplifted by fleet on total monthly basis
-On a good fuel forecast a standard error of +/- 3 percent is considered within tolerance.
6) If new fleet type normalize estimated fuel burn performance numbers to ASM analysis and apply the ARIMA forecast to the numbers and future schedule.
7) The delta between current fleet fuel burn and predicated fuel burn of the revised fleet.
8) Include that analysis in conjunction with NPV, revenue management, crewing models.
For Fleet /Schedule Expansion,
1) Same as above except modify the future schedule and run fuel burns for proposed fleet types.
Operationally fuel forecasting is done on a regular basis and consistently error checked. Fuel forecasts are done with a new aircraft type, monthly for budgeting, fuel purchasing, fleet acquisition and as an as needed basis. The forecasts are highly reliable and a critical component to strategic and operational decision making. It is surprising but fuel data/asm from historical data is all you need to build a high quality forecast. It sounds counter intuitive but these forecasts are just as effective as getting into the details. The only unknown when applying it to operational evaluations is the estimated fuel burns from the suppliers for a new type but again that data is usually fairly conservative. I wish I could comment on the A330-900 versus 787-900 analysis specifically. Only thing I can say is that its been done. The above is a fairly brief version of how fuel burn is included in an aircraft evaluation process.