Start here
Business Intelligence (BI) in plain language
A sales chart becomes useful when people agree what counts as a sale, which period it belongs to and whether returns are deducted. BI connects these definitions to data that users can inspect.
Reports describe patterns and support investigation. A change in a chart does not by itself explain why it happened or prove that an action caused it.
How it works
Start with the decision
Identify the question, responsible user and action the report should inform.
Define and reconcile measures
Specify numerator, denominator, time period and exclusions. Check totals against the source.
Present context
Show filters, units, refresh time and useful comparisons alongside the result.
Illustrative example · Fictional data
Comparing two fictional sales channels
- Input
- Channel A has 20 orders from 100 enquiries; channel B has 30 orders from 300 enquiries in the same period.
- Process
- Calculate conversion rates as orders divided by enquiries: A is 20%; B is 10%. Compare volume and rate rather than choosing only one.
- Expected output
- B produces more orders, while A converts a larger share of enquiries.
- How to check it
- Confirm both channels use the same enquiry definition. These figures do not establish profitability or explain the difference.
Where it helps
- Tracking agreed operational measures.
- Comparing periods or business segments.
- Finding exceptions that warrant investigation.
Report, BI and predictive analytics
- Report
- A presentation of selected information.
- BI
- The wider practice connecting definitions, data and reporting to decisions.
- Prediction
- Estimates a future or unknown outcome and needs separate validation.
The five layers of a business-intelligence stack
Sales transactions, customer interactions, financial records and supply-chain events begin in different systems. A BI stack connects those records to questions such as which products are profitable, which customers are leaving and where operations are slowing down. A dashboard is the visible end of that process, not the whole system.
- Sources: operational databases, CRM, ERP and cloud applications supply the underlying records.
- Integration: ETL or ELT pipelines align formats, keys and update schedules across those sources.
- Storage: a warehouse holds integrated historical data for analysis rather than individual operational transactions.
- Semantic layer: shared definitions specify revenue, active customers, reporting periods and other business measures so reports do not silently disagree.
- Presentation: tools such as Power BI, Tableau or Looker expose reports, filters and drill-down paths for business users.
Self-service BI still needs shared definitions
Self-service BI lets an analyst, finance manager or operations lead build and modify reports without requesting every change from a central reporting team. It does not remove the need for IT and data teams: access, refresh, trusted datasets and metric definitions still need owners.
Separate operational reporting, such as today’s order backlog, from strategic reporting, such as quarterly margin trends. A leading indicator helps a team anticipate a result; a lagging indicator describes an outcome already recorded. Name the period, population and calculation for each KPI before comparing teams or publishing a scorecard.
BI commonly supports descriptive and diagnostic questions: what happened and why. Analytics can also include forecasting and optimisation. The boundary is not absolute; a dashboard can display a forecast produced elsewhere without itself being the forecasting model.
Power BI or Tableau: what should you compare?
Start with the systems your organisation already supports, the analyses users need and who will maintain the reports. Evaluate source connectivity, data modelling, visual exploration, access control, refresh and distribution with the same representative dataset.
Compare the actual licences and sharing setup rather than assuming every capability comes with an existing office subscription. A successful pilot should show that two users get consistent figures and the right permissions, not only that one author can create an attractive chart.
Related reading and learning pathways
Common pitfalls
- Filters can make apparently identical metrics disagree.
- Averages can hide important segment differences.
- A recent-looking dashboard can contain stale source data.
Put the idea to work
Try it yourself
Recalculate the channel example after adding ten cancelled orders to B. Decide whether your measure counts placed orders or completed orders and label it accordingly.
Common questions
Is BI the same as Power BI?
No. BI is the practice; Power BI is one product used to implement it. Other BI tools include Tableau, Looker and Qlik. The practice covers collecting, modelling and presenting data so people can make better business decisions. Tool choice matters less than clean, well-modelled data.
Does a dashboard explain causation?
No. It can reveal patterns, but explaining causes requires additional evidence and analysis. A dashboard may show that sales fell after a price change, but other factors could explain it. Experiments, segmentation or statistical analysis help test the cause.
Your next step
Continue learning
Connected concepts
Sources and further reading
Examples on this page are teaching exercises using fictional information. They are not client case studies or measured outcomes.