Major projects often deviate from their expected budget, schedule, and benefits. Existing techniques to improve the forecasted outcome, such as the Reference Class Forecasting technique, ignore any flexibility in the project timing. This paper acknowledges this flexibility and proposes a real options model to optimise the project timing and value when the project investment cost, the construction schedule, and the output value are uncertain. We illustrate the approach with thin and fat upper tails distributions of cost and schedule overruns. Finally, we show how practitioners can implement the proposed approach using empirical distributions of costs and schedules from comparable prior projects.