The British papers are full of stories about The Great British Rail Network Franchise Disaster of 2012.  Like Bristow’s Great Tea Trolley Disaster of 1967, we may never learn the real reasons behind the disaster — errors are alleged in calculations (arithmetic errors? using multi-line spreadsheets?) undertaken by senior civil servants, now suspended.  But one item leapt out to me:

Government sources said “heads will definitely roll in the department” over the affair, adding that “the minister cannot be expected to be responsible for a very technical models with hundreds of lines in a spreadsheet”.
The key error seems to have been to underestimate the potential value of the franchise – where the company pays a premium to the Government, rather than receiving a subsidy.
The department said mistakes had been made over estimates of the number of passengers who would use the route and the way inflation was calculated. Three civil servants have been suspended.

Why on earth are government civil servants estimating future passenger numbers and rates of inflation?  Surely, that is the business of the bidders.   Only the bidders, after all, have the expertise, the experience, and the motivated self-interest to make these forecasts as accurately and realistically as possible.  The Government should be making its franchise decision on whatever criteria it thinks appropriate (eg, the numbers of jobs created, the novelty of services provided, the public fares charged, the money payments offered for the franchise, etc), but not trying to second-guess the business plans of the train operators.   Any demand forecast will depend on assumptions about the actual services offered, the actual prices charged for these services, and the actions undertaken to market, promote, distribute, and sell them, and none of these assumptions are within the purview of the Government.
Indeed, not only do civil servants not know these marketing plans, civil servants — in my extensive experience of submitting telecommunications licence applications — do not even have the expertise needed to assess such plans.     How can they tell whether a marketing plan is effective or not?  Feasible or not?  Sensible or not?  Even experienced marketers can get market planning wrong, so how much more so civil servants with no commercial experience at all, no direct stake in the outcome, and no ear to the market ground?  A famous British example of marketing ignorance by civil servants was the refusal by British Treasury officials during the 1960s to approve (what is now) British Telecom’s proposed telecoms switch upgrades, since the proposed switches allowed for itemized billing of calls:  What user would need that? asked the refusenik officials.
A decade of telecommunications licences awarded by beauty contests finally convinced Governments around the world to put aside any attempt to plan the businesses involved, and just ask potential operators to pay what they think each licence is worth, via auctions.   Of course, British regional rail network franchises are monopolies, so it is appropriate for franchise allocation decisions to be based on criteria additional to the amount of money offered for the franchise.   It is even appropriate for these criteria to include subjective and qualitative factors, such as the degree of risk of the bidder going bankrupt during the franchise period.  Even so, I cannot see a need for a Government to be predicting customer demand,  or even assessing the predictions of customer demand made by the bidders.  They should leave that job to the people with the most to lose for getting the forecasts wrong.
If, for some reason, the Government does need its own independent forecast of demand, it should outsource the creation of the forecast (strictly, the forecast model) to some outside entity with the expertise, the experience, and the motivated self-interest to make these forecasts (or model) as accurately and realistically as possible.   Outsourcing would also more likely ensure that the generation of such demand forecasts is independent from their use in any evaluation of franchise bids, so that neither decision — deciding the forecasts nor choosing the franchise winners — could corruptly influence the other.