Weighted Shortest Job First
Sequencing work by dividing the cost of delaying it by the effort to deliver it, which maximises value delivered per unit of time.
Prioritisation by value alone sequences large valuable items ahead of small valuable ones and delivers less total value. Dividing by duration corrects that, and the result is provably optimal for a queue of independent items.
Cost of delay ÷ job size. Cost of delay combines the business value, the time criticality — does the value decay, is there a deadline or a market window — and the risk reduction or opportunity it unlocks.
The technique's real contribution is not the arithmetic. It is that estimating cost of delay forces a conversation the organisation avoids: what does it actually cost us to not have this for another quarter? Many items turn out to have a cost of delay close to zero, which settles their priority without further debate.
For architecture and platform work specifically, this is the framing that makes the case competitive with features, because enabling work has a compounding cost of delay — every quarter without it, every subsequent item is slower. Stated that way it competes on the same axis rather than being classified as "technical" and deferred indefinitely.
Two cautions: it assumes items are independent, so dependencies must be resolved first; and estimates are judgements, so it is best used to produce relative ordering and surface disagreement, not to generate a precise ranking that is then defended.