"We need better analytics" is one of the most common requests in business technology. It is also one of the most imprecise. Analytics — the practice of using data to generate insights — spans a spectrum from basic operational reporting to advanced machine learning, and conflating different parts of this spectrum leads to misaligned investments and unmet expectations.
Business Intelligence (BI) answers the question: what happened? BI tools — Tableau, Power BI, Looker, Metabase — connect to data sources, provide visualisation capabilities, and enable business users to explore historical data through dashboards and reports. A well-implemented BI system gives every manager and executive real-time visibility into the metrics relevant to their function. BI is not sexy but it is foundational, and most organisations still have significant headroom to improve their BI capability.
Descriptive analytics is the formal name for what BI does: describing historical data. Diagnostic analytics adds the next layer: why did it happen? Root cause analysis tools, drilldowns, correlation analysis, and attribution models fall into this category.
Predictive analytics answers what will happen — using historical patterns to forecast future outcomes. Credit scoring, demand forecasting, churn prediction, and equipment failure prediction are predictive analytics applications. They require ML models, not just SQL queries, and they require a data science team to build and maintain them.
Prescriptive analytics answers what should we do — optimising decisions based on predicted outcomes. Optimal pricing, supply chain optimisation, marketing spend allocation, and personalised recommendation are prescriptive analytics applications. These are the most technically complex and often the most commercially valuable.
Most organisations benefit from investing in BI and descriptive analytics first, not because advanced analytics is not valuable but because the data foundations, data literacy, and organisational decision-making processes needed to use advanced analytics do not exist until BI has been established. Advanced analytics built on poor data quality and in organisations that do not make data-driven decisions in their basic operations will not deliver value regardless of the sophistication of the models.