Business intelligence (BI) dashboards are screens of charts and tables that an analyst builds once from a company's data, and that refresh themselves with new numbers every day or every week. A dashboard is read again and again, by many people, often in the same meeting each week. Each reader sees only the numbers that he or she is allowed to see.
A dashboard does not store the data itself. It sends queries to the company's data warehouse and draws the answers as charts. A query is a written question in SQL, the language that databases understand. Some BI tools also pull data straight out of other apps through built-in data connectors, which are ready-made code for each app. The tables are often prepared first, by data engineers or by an analyst's analytics automation workflow.
Take, for example, the school education department of a state government, with its head office in Jaipur. It tracks three numbers for each of the state's 60,000 government schools: student attendance, midday meals served and teachers present. Each school sends these figures from a phone app every afternoon, and they are copied into the department's warehouse overnight.
An analyst in the department builds one dashboard. It has a map of the state's districts, coloured by attendance. It has a chart of meals served each day, and a table of the 50 schools with the lowest attendance. Each chart is defined once: which table it reads, which number it adds up and how it splits that number, such as by district or by week. At 6 a.m. every day, the dashboard runs its queries again. Nobody redraws the charts.
The secretary of the department sees the whole state. A district education officer in Barmer opens the same dashboard but sees only Barmer's schools. This is called row-level security. The dashboard adds a condition to each of her queries, such as "district = Barmer", based on her login. So the department builds one dashboard, not one for each district. A reader can also click on a district to see its blocks, and then click on a block to see its schools. This is called drilling down.
There are two common ways to buy BI. Enterprise BI is built for large organisations, where thousands of employees read the same approved reports. Cloud BI runs as a web service, a kind of SaaS, so the buyer installs nothing.
A dashboard is only as correct as the data under it. If last night's copy failed, the chart shows old numbers with no warning. So many teams also run data observability tools, which raise an alert when a table arrives late or incomplete.
A dashboard answers the questions that an analyst planned in advance. It cannot answer a new one. When a reader asks something that no chart shows, the question goes to an analyst, or to a tool for asking the data in plain language. A dashboard is also different from embedded reporting, which puts reports inside another company's product, for that product's own customers.