Is forward testing necessary after a successful backtest?

Historical testing can provide evidence of potential, but real-time conditions introduce uncertainty about execution and changing behaviour. Forward testing allows a strategy to be observed on new data before greater reliance is placed on it. How long would you forward-test a strategy before considering it ready for an evaluation?