From Concepts to KPIs

How does a business calculation become a measure in Veezoo? This lecture builds three example calculations, all in natural language. First a row-level calculation, the number of days each promotion ran, computed from its start and end date. Then an aggregate calculation, the price premium percentage, how much higher the suggested price is than what is charged today. Finally a nested calculation that reuses that new measure and scales it by how confident each recommendation is.

Along the way the lecture shows where each kind of calculation lives, how Veezoo tests what it builds, and how a saved measure appears in the Knowledge Graph.

Key points

• Three levels: every calculation is either row-level, aggregate, or nested.
• Natural language is enough: describe the calculation in chat, and Veezoo builds it, tests it, and flags any similar measure that already exists.
• Row-level: computes a value per record and lives in the sql property, for example promotion duration from end date minus start date.
• Row-level values can still be aggregated on top, for example the average promotion duration across promotions.
• Aggregate: combines aggregated numbers within one class, for example average suggested price against total sell price as a percentage.
• Saving an aggregate calculation stores it as a definition with no sql property, and the measure appears in the Knowledge Graph.
• Nested: builds a new measure on top of an existing one, for example the price premium percentage times the confidence score.

Check your understanding