I had the occasion to discuss the topic recently and I realized that there's no "consensual truth" about it. They are basic Business Intelligence concepts but I realized They are still debated. Nobody ever cares about the obvious so basic concepts like these give the rise to complex discussions.
What I find amazing is the amount of "it is kinda like…" "it's more focused on… " "it depends…" definitions I've heard. Maybe I'm too naive, but the idea is clear and simple, in my opinion.
A measure is a quantity attached to a fact. The quantity and the value of an invoice row are measures. The number of active customers is a measure. A single transaction value on the ledger accounts is a measure. The tag price of a product in line of purchase is a measure. The type of aggregation and the summary level of the related facts is not relevant for the definition and may vary. Please note that running or plain totals are simply an aggregation method and they're not relevant either.
A derived measure is simply the result of a mathematical operation on simple measures.
A metric is a measure sampled over time. In other words a measure presented by showing how its value changes over time is a metric. There is the typical display of gauges with a slider which shows the metric over time but an excel table with weeks and the invoiced value per week is a good metric representation. A graph with quantities on the y axis and time on the x axis is a metric visualization either. Often derived measures are used as metrics, but it's not a rule at all.
While the measure has no privileged dimension for analysis, a metric is relevant over time first, and over other dimensions in the second place.
Some might think that, from a strictly technical point of view there's no substantial difference in presenting a measure or a metric and the transition between the two is seamless; they're right.
A KPI (Key Performance Indicator) is a metric which is included in a performance management system. That is, a metric with associated targets or oscillation boundaries or any other sort of constraint is a KPI. In a BI system there may be thousands of measures and metrics, but few tenths of KPIs.
The "Key" part of the term refers to the particular relevancy of the metric for the health of the organization. The "Performance" part of the term refers to the associated values that tell us if the metric value and/or trend are good or bad. The "Indicator" part of the name stresses that the metric chosen should be influenced by other metrics and measures and should give an overall view of an organizational segment.
Choosing the KPIs for a performance management system, for example according to the Balanced Scorecard model, is not a technical issue, but a large business project involving the management.
I hope I've been clear enough. Let me know what you think of it and, if you come up with a clearer definition, I need to know!
Enjoy
