
Seven families of analytics, and two actions that decide whether the output changes a decision or just decorates a slide.
Summarised from the Go2Cab business insight series
This sits beyond the target operating model, where the customer, the value chain, the services and the capabilities that deliver them are defined. Go2Cab applies analytics across seven areas — the same discipline whether the data sits in Dynamics 365, Business Central or an operational system of record.
Start with a few measurable, testable objectives that directly affect customer needs, then check whether meeting them will delight the customer or simply be expected. “Measurable” decides what data is needed, how often, and within what margin of accuracy — which in turn decides the collection mechanism. “Testable” reaches further: how the operation is run over time, the timing and specificity of notifications, and the actions that sustain the optimal level.
Verification and validation are different tests. Verification proves existence or absence — is there a missing objective? Validation proves the outcome in real or near-real time, including the sequence of delivery. A lift that closes its doors before announcing “going up” passes verification and fails validation.
Assuming the raw data is sound, the temptation is to present the evidence as an average, a percentage or a tally chart — churn, response time, incidents cleared, pipeline, downtime, provisioned capacity. The height of a bar or the area of a pie sector is not a decision. A larger bar can mean more sales or more defects.
Percentages mislead in the same way. Eighty percent of customers being satisfied does not mean twenty percent are content, and it does not establish that eighty is high enough. Without the variation, the sequence and the operational dynamics, the number cannot tell you what to do next.
Analytics only changes outcomes where it is embedded in the flow of work. In Microsoft Dynamics 365 that means predictive measures on finance operations — invoicing, accounts payable, three-way matching — and on supply chain, where just-in-time performance depends on forecast accuracy and exception detection rather than monthly averages.
Business Central gets the same treatment at mid-market scale. And Microsoft Copilot becomes useful once the measures it reads are validated: an AI assistant answering from a verified measurement layer speeds up decisions, while the same assistant answering from aggregated averages merely accelerates the wrong one.
The takeaway
Define a measurable, testable objective before choosing the algorithm — then validate it in near-real time, not once at the end of the project.