In operational flood forecasting, the decision to issue a warning to the community at risk has to be made in real time based on evolving probabilistic forecasts of the hazard. However, not only is the hazard forecast changing in time, but so too are the exposure and vulnerability. This may be the result of changes in the location of assets and people, or - more qualitatively - in a loss or gain for the reputation of the forecasting authority. In this work, we take the first steps towards quantifying these to form a risk-based framework for making the decision to issue a warning using an example catchment in the UK.