Business intelligence (BI) is the software that a company uses to look at its own data and get answers from it: the reports, the dashboards and the charts on the screen in a Monday meeting. Analytics is the broader word. It also covers working out why a number moved, and estimating what it will be next month.
Two groups of people use this software. Analysts are employees whose job is to work with data. They prepare it, build the dashboards and answer questions from the rest of the company. Managers read the dashboards and ask the questions. The data itself usually sits in a data warehouse, one large store into which the company copies the records from all its other systems. The page on data engineering explains how the copies get there.
Consider the head of sales at a packaged-foods company in Kolkata that sells biscuits and snacks through 300 distributors. Distributors are wholesale firms that buy from the company and supply the neighbourhood shops.
On Sunday night, an analytics automation workflow runs. An analyst built it once from ready-made steps. It pulls the week's orders, cleans them and adds them up by state, product and distributor. At 9 a.m. on Monday, the head of sales opens the BI dashboard. It shows last week's sales by city. Guwahati is down 18%, and every other city is up. Nobody rebuilt the dashboard for this meeting. It was built once, and it refreshes itself every week. The distributors see the same week's numbers too. Each one sees its own orders and dues inside the distributor management system, whose reports come from an embedded reporting engine built into that software.
The dashboard shows what happened. It does not show why. So the head of sales types a question: "Why did Guwahati fall last week?" A tool for asking the data in plain language turns the question into a database query, runs it and replies. Two large distributors in Guwahati placed no orders, because heavy rain closed a highway and held up their trucks. A tool that answers in this way is called an AI data analyst.
Next, an analyst checks whether last month's discount on the ₹10 pack really raised sales, or whether sales would have risen anyway. She runs this test in statistical software. Finally, she plans for Diwali. A time-series forecasting model reads three years of weekly sales, which are a kind of time series data. It estimates how many cartons each state will need in the six weeks before the festival. These numbers go to the supply-chain planning team, which decides what the factories make.
The three tools that most companies use for dashboards are Tableau, Power BI and Looker. Newer companies are building software in which a typed question takes the place of the dashboard. Whether the question replaces the dashboard, or sits beside it, is still an open question.