This research aims at developing an intra-firm prediction market system as a tool for a quantitative forecasting problem, such as project lead-time estimation, demand forecasting, etc. The system uses the variable-interval prediction security (VIPS) as the prediction security to be traded in the market, and it possesses a computerized market maker that evaluates each unit of VIPS with a Gaussian price density and updates the price density function through a LMSR-like updating logic in an intermittent updating frequency according to the transactions in the market. Laboratory experiments on a virtual demand forecasting problem are conducted to study how the prediction market system functions with only a small number of participants. The system provides satisfactory results as compared to a statistical method, provided that the parameter controlling the sensitivity of the updating logic is appropriately tuned.