Attribution coverage is the measure behind the AI Value Gap: the gap is the share of AI spend that is not connected to a measured outcome, and coverage is the share that is.
Computing a first version takes three steps. Inventory all AI spend, including seats and AI tiers embedded in SaaS contracts, not only API and token spend. Classify each line as outcome-linked (a measured result exists, with a baseline and an owner), activity-linked (usage data only) or unattributed. Then publish the ratio.
The first figure is allowed to be low. There is no reliable industry average to compare it with yet, so the useful signals are the direction of travel and the quality of the evidence behind it. A coverage figure that quietly divides by allocated spend rather than all identified spend flatters the result. Being explicit about the test matters too: the same portfolio can show one figure on "has a baseline and an outcome measure" and a lower one on decision-grade evidence, as the evidence levels explain.
Attribution coverage is not the same as scrap rate, a measure from the Tokenomics Foundation's draft AI value data layer. Scrap rate asks how much of the bill was wasted in production on failed, abandoned or untied work. Attribution coverage asks how much of the bill has evidence that it was worth paying. A workload can score well on one and badly on the other, so report the two side by side.
