Business intelligence

01The problem

Every Monday, Kolanut's sales manager asks for the same numbers: last week's revenue, revenue by region, overdue customers. Every Monday an analyst spends two hours rebuilding the same spreadsheet. The numbers are useful; the process is wasteful, and each rebuild risks a new mistake.

Business intelligence (BI) fixes this: build the analysis once, connect it to the data, and let it refresh.

02The concept

Analytics and BI

The words overlap, but a useful distinction:

Good analysis often becomes BI. Once you know North West matters, you put it on the dashboard so nobody has to ask again.

The BI pipeline

1. Sources: the systems where work happens: the order system, billing, HR, CRM, and the odd spreadsheet.
2. ETL (extract, transform, load): copying data out, cleaning it and shaping it. In Power BI this is Power Query. Every cleaning step from lesson 5 becomes an automated step that runs on each refresh.
3. Model / warehouse: clean tables with relationships (lesson 4) and agreed calculations: a single source of truth. A large company keeps this in a data warehouse, a database built for analysis; a small one may keep it inside the BI file itself.
4. Dashboards and reports: the views people look at, refreshed on a schedule.

A single source of truth

Without BI, every team builds its own spreadsheet, and their numbers disagree. Sales says revenue was ₦48m (before discounts), finance says ₦46m (after discounts and returns), and the meeting is spent arguing about whose number is right.

A single source of truth means one place where each KPI is calculated, one way, from one set of cleaned data. Everyone's dashboard reads from it, so everyone sees the same number. The definitions from lesson 2 (name, formula, inclusions, source) are written into the model once.

Dashboards and reports

Most BI tools produce both, and a good dashboard links through to the report for detail.

Designing a dashboard

1. Start from the decisions. Who opens this, and what will they do differently because of it?
2. Top-left is read first. Put the headline KPIs there.
3. Every number needs a comparison: against last year, last month or target. "₦290.7m" means little; "₦290.7m, up 19% on last year" means something.
4. Use the chart rules from lesson 7. Lines for trends, sorted bars for comparisons, one highlight colour.
5. End with an action list: the customers to call, the invoices to chase.
6. Leave things out. Four KPIs and three visuals beat twenty tiles nobody reads.

Self-service and governance

Modern BI tools let business users build their own reports from the shared model: self-service BI. It's powerful, but it needs governance:

• Certified data: the official model is marked as trusted, so people build on it rather than on private copies.
• Access: each person sees the data they're allowed to; a regional manager might see only their region. In Power BI this is called row-level security.
• Ownership: each dashboard and each KPI has an owner who answers questions about it.

Common BI tools

The ideas in this lesson apply to all of them; the menus differ.

03Example

A sensible first dashboard for Kolanut's sales manager:

Four KPIs, three visuals, one action list. The discipline is leaving things out.

Note the leading indicator in the bottom table: a customer ordering less often is an early warning, before revenue falls. That's exactly what happened in the North West, and a list like this would have flagged those shops while there was still time to act.

04Walkthrough

When a law firm like Ashgrove Chambers builds BI for its partners, it follows the same steps:

1. Agree the KPIs with the partners: open matters, hearings adjourned, invoices overdue, collection rate.
2. Define each one precisely. Collection rate = paid invoices ÷ all invoices issued × 100, counted by number of invoices. That choice matters: counted by value instead, Ashgrove's rate is 81.5%, because paid invoices are larger on average than unpaid ones. Both are reasonable; what's not reasonable is two partners using different ones.
3. Connect the sources: the matter management system and the billing system.
4. Check the numbers against a manual calculation before anyone relies on the dashboard.
5. Schedule the refresh, for example every morning at 7.
6. Name an owner for each KPI, who answers when someone asks "why did this change?"

Step 4 is the one people skip. A dashboard that is wrong once loses trust for months.

Summary

05Practice

06More practice

Optional drills. They don't count towards the certificate, but they're the fastest way to make this lesson stick. Several use a different dataset from the lesson on purpose: if you can do the same thing on unfamiliar data, you've really learned it.

07Check your understanding