Methodology
Having data is not having enough data
A margin computed over the products that happen to carry a recorded cost is not the store's margin. Coverage is how Autopsify measures the gap between the two — and how it decides which conclusions the evidence can actually carry.
Operator-led · Read-only Shopify access
The idea
Coverage is a property of the measurement
Coverage asks a narrow question: of the thing being measured, how much did the evidence actually reach? If cost data exists for part of the catalogue, any margin figure describes that part and not the rest. The figure is not wrong — it is narrower than it looks, and the narrowness is invisible unless someone states it.
This matters most when coverage is uneven. A catalogue where the missing costs are concentrated in the highest-volume products produces a margin figure that describes almost none of the revenue, while still appearing to describe the store.
What is measured
The coverage checks
Cost coverage
The share of revenue carrying a recorded unit cost, weighted by revenue rather than by product count. Weighting matters: a catalogue can have costs on most products and still be missing them on the few that generate the revenue.
Advertising data coverage
Whether spend is present for the whole period from the connected accounts, and whether the connection was live throughout it. A lapse mid-period understates spend without producing any obvious error.
Attribution coverage
The share of counted revenue that carries a recognisable marketing source. Below roughly half, every channel-level acquisition figure is a floor rather than a measurement, and no channel-level conclusion can be strengthened in either direction.
Cost completeness
Which cost categories exist at all. Cost of goods may be complete while payment fees, fulfilment and overhead are entirely absent — which is why the ladder frequently stops at contribution rather than profit.
Date-range alignment
Whether the periods being compared are genuinely comparable: the same length, the same costing basis, and both fully covered by the data. A comparison against a period with thinner data measures the data, not the business.
Source availability
Whether a dataset is reachable at all. Some Shopify data is restricted or simply not exposed, and some figures — conversion rate from sessions, for instance — cannot be computed for that reason rather than because of anything the merchant did.
Reconciliation
Whether the figures agree with each other. Deductions that do not reconcile against the totals they came from indicate a measurement problem before they indicate a business problem.
The distinction
Coverage is not performance
This is the confusion worth naming explicitly, because it cuts both ways.
A store can be performing well with incomplete measurement. Thin cost data does not make a healthy business unhealthy; it makes the health harder to demonstrate. Equally, a store can be performing badly with complete measurement — every figure available, every one of them poor.
Coverage determines the strength of the conclusion, not the quality of the business. Low coverage is a reason to be careful about what is claimed, and often a reason to go and fix the measurement first. It is never, by itself, a finding about performance.
The consequence
Why a figure gets withheld
When coverage is insufficient, the Blueprint has three options: report the figure as though it were complete, report it with a caveat nobody reads, or withhold it and say why. The method takes the third.
Withholding is not the end of the analysis. The report states which input is missing, what the figure would require, and — where the gap is fixable — treats closing it as an action in its own right. Improving coverage changes what can be measured next cycle, which is frequently more valuable than the figure that could not be produced this cycle.
Coverage is measured against the evidence classification and applies to every rung of the economic reconstruction. It also governs comparison: a re-diagnosis is only meaningful if coverage held steady between the two readings, because a margin that moved because a different set of products became costed is not a margin that moved.
Methodology
The rest of the method
- The methodHow the whole diagnostic fits together.
- Economic reconstructionRebuilding the ladder from gross sales to whatever the evidence reaches.
- Evidence modelKnown, Supported, Unknown — and why Unknown is never zero.
- Verification & re-diagnosisWhat the reading about 30 days later can and cannot prove.
Find out what your data can actually support
Readiness is checked before payment. If coverage is too thin for the analysis to be worth anything, we tell you that instead of selling you a cycle.