A proposed model would improve prediction accuracy from 91 per cent to 93 per cent compared with a simple business rule already in place. Running it would cost roughly eighty times more per prediction. Which consideration should carry the most weight in the decision?

• AThe two point gain in accuracy, since accuracy is the primary measure of a model.
• BThe F1 score of the new model compared with the rule.
• CThe business return, weighing what two points of accuracy is worth against cost per prediction and development cost.
• DWhether a more complex architecture could push accuracy higher still.