Diagnostic data: test their decision power, not their volume.
Answer:
A diagnosis does not require all available data. It requires data able to distinguish the options actually under consideration. Aggregated sales, media spend and list prices are insufficient when the decision concerns segment margin, channel costs or transfers between products. Start from the decisions, define the required grain, then test coverage, quality, variation, freshness and comparability. High volume at the wrong grain creates apparent precision without discriminating power.
Decision:
Allow a diagnosis only when critical data clear their non-compensatory thresholds, or when an uncertainty plan explicitly bounds the gaps.
Principles:
The grain is product, segment, channel, geography and period. A national monthly margin cannot arbitrate a regional weekly channel. Exact but constant data cannot distinguish a response: price, availability, exposure or intensity need sufficient, non-perfectly-synchronised variation across the future decision range. Definitions for revenue, active customer, order, return and media spend require source, rule, unit, time zone, cancellation treatment and break date.
Method:
List the options and decision criteria, then derive economic outcomes, levers, context, constraints and joining identifiers. Score each data domain from 0 to 100 for coverage, quality, variation, freshness and comparability; for critical data use the minimum, not an average. For every gap estimate the uncertainty it adds and its chance of changing the decision. Correct the source, narrow scope, use an interval, collect targeted data or postpone according to financial exposure and reversibility.
Worked example:
An omnichannel diagnosis has sales, margin, media, availability, price-promotion and returns. Average scores exceed 70, but channel-level margin coverage is 42 because logistics and returns costs are pooled at headquarters. Critical margin score = min(coverage 42, quality 88, variation 76, freshness 95, comparability 61) = 42. After activity-based cost allocation, direct-channel contribution falls from +€14 to +€5, with an interval from €1 to €9. The approved expansion threshold is €7; estimated probability of clearing it is 35%.
Conclusion and limits:
The overall average does not justify expansion. The diagnosis may continue for other interfaces, but the channel decision is suspended until cost allocation is available or converted into a limited pilot. A readiness score does not detect every selection or causal bias; thresholds depend on financial exposure; historical quality does not guarantee stable future mechanisms.