Choosing the right metrics is the foundation of useful BI. Bad metrics to bad decisions even with accurate data. Let us cover: 1) Properties of a GOOD metric: — TIED TO A GOAL: a metric should reflect what really matters for the business/decision, NOT what is just easy to measure. Ask: what goal/question does this metric reflect? If none specific — why is it there? (a link to "start with a question" from analytics). — CLEAR AND UNAMBIGUOUS: it is clear what the metric means and HOW it is computed. If different people understand "an active user" or "revenue" differently — the metric misleads. Define metrics clearly (what is included, over what period, how computed). — ACTIONABLE: you can INFLUENCE the metric and make DECISIONS on it. "A number we do not influence and decide nothing on" is useless on a dashboard. A good metric is tied to actions. — HONEST and non-distorting: reflects reality, does not create a false picture. 2) TYPICAL metric ERRORS (important!): — "VANITY METRICS": nice growing numbers NOT tied to real value/decisions. E.g. "total registered users" keeps growing (cumulative), looks good, but does not say whether things are going well NOW (how many are ACTIVE? paying?). Vanity metrics create a false sense of success. Choose metrics reflecting real value, not just pleasant numbers. — The MEASURABLE instead of the IMPORTANT: measuring what is easy to compute instead of what matters (but is harder). E.g. counting "the number of calls" (easy) rather than "resolved customer problems" (important but harder). Do not substitute the important with the convenient. — ONE METRIC WITHOUT CONTEXT: one metric can mislead without accompanying ones. Revenue growth is good, but not at any cost (what about expenses? margin? churn?). Often a BALANCE of metrics is needed (do not optimize one at the expense of others). — AVERAGES HIDING DIFFERENCES: an average metric can mask problems in segments (from the analytics course — averages hide). — WRONG INCENTIVES (Goodharts law): "when a metric becomes a target, it ceases to be a good metric". If people know they are assessed by a metric, they start optimizing IT, sometimes at the expense of the real goal. Example: a metric "number of closed tickets" to they close tickets fast but do not solve problems. Choosing a KPI, think what BEHAVIOR the metric incentivizes, does it not create harmful incentives. 3) KPIs (key metrics): — Of all metrics, the KEY ones (KPIs) are singled out — a few most important reflecting the main goals. Do not turn a dashboard into a dump of 50 metrics — single out the key ones that are really tracked and decided on. — Fewer but right metrics are better than many random ones. 4) The role of AI: — AI helps COMPUTE metrics (queries/code from a description), BUILD them on dashboards, explain. — AI can SUGGEST possible metrics for your task. — But the CHOICE of the right metrics is with you: only you know the real goals, context, what decisions you make, what behavior you want to incentivize. AI will compute any metric, including a useless/harmful one; WHAT to measure and why is your decision (a link to "you pose the question" from analytics). 5) The practical approach: — Start with GOALS/decisions: what matters for the business? what decisions do we make? — Choose metrics REFLECTING this (tied to a goal, actionable), not just easily computed or pleasant. — Define them CLEARLY (what is included, how computed). — Check for traps: are they vanity? do they substitute the important with the convenient? do they create harmful incentives? is a balance with others needed? — Single out KPIs (a few key ones). — Use AI to compute and show the CHOSEN metrics, but the choice is with you. Practical rule: the right metrics are the foundation of BI; bad ones lead to bad decisions even with accurate data. A good metric is TIED TO A GOAL (reflects the important, not just the easily computed), CLEAR (clearly defined), ACTIONABLE (we influence and decide) and HONEST. Avoid the traps: vanity metrics (pleasant but not about value), the measurable instead of the important, one metric without balance, averages hiding differences, harmful incentives (a metric as a target distorts behavior — Goodharts law). Single out a few KPIs, not a dump of metrics. AI computes and shows any metrics, but the CHOICE of the right ones (tied to goals, honest, without harmful incentives) is with you, knowing the context. What you measure determines what the team looks at and how it acts — so the metric choice is critical.