Cost allocation
Allocation fails on the
spend nobody claims
Tag coverage counted in resources flatters you; counted in dollars it tells the truth. Zephto measures coverage by spend, names the unallocated remainder instead of spreading it, and reports AI consumption against the feature that burned it.
The short answer
How allocation is measured
Zephto measures tag and label coverage as a share of spend, not as a share of resource count, and reports unallocated cost as its own named line rather than distributing it by assumption.
For AI workloads it goes further: token and DBU consumption is attributed to the deployment, job or workspace that produced it, which is the only level at which cost-per-feature is a real number rather than an allocation key.
- FinOps practitioners
- Product and platform owners
- AI engineering leads
- Finance business partners
Why coverage numbers mislead
Resources tagged is not spend allocated
Ninety per cent of resources tagged sounds like a solved problem. It routinely means sixty per cent of spend allocated, because the untagged remainder is where the expensive things live — the shared capacity, the platform subscription, the workspace nobody owns.
Spreading that remainder pro-rata makes the report add up and makes every number in it unfalsifiable. Naming it keeps the report honest and gives someone a reason to fix the tagging, which is the only thing that actually improves the answer.
AI spend makes this sharper. A single Foundry deployment can serve four product features, and the bill knows nothing about any of them. Attribution has to come from deployment naming, workspace structure and tagging discipline — which is why untagged token spend is a governance finding even though it costs nothing extra by itself.
Find out what share of your bill is actually allocated
The audit reports coverage in dollars, and names what is left over.