I continue to draw on my years of experience in building machine learning products, by breaking the process down into a framework of eight key steps. Here I cover steps four through to six; exploring the cost of getting things wrong, to building a feedback loop.