Knowledge base

Glossary

81 terms from the world of business intelligence, explained in plain language. Written for the person in the meeting, not the person building it.

Fundamentals

14 terms

Analytics #: The broader term for finding patterns in data, ranging from simple counts to predictive models. Business intelligence is analytics applied to your own company figures, packaged into dashboards everyone understands.
Business intelligence BI #: Turning your own business data into figures you decide on. Not a one-off report but something that keeps running: figures arrive by themselves and mean the same to everyone.
Dashboard #: One screen with the key figures for your business or department, updating itself. You see how you are doing in seconds, then click through for the why.
Data analysis #: Examining figures to answer a specific question, such as why revenue dropped in March. It is a one-off investigation; business intelligence is the system that keeps delivering such insights continuously.
Data maturity #: How far your organisation has come in using data, from collecting by hand to correcting course automatically. It decides what a sensible next step is.
Data strategy #: The plan that sets out which data sources, systems and skills your organisation needs to reach its goals. Without a data strategy you build dashboards by chance instead of by route.
Data-driven working #: Deciding on measured facts instead of instinct or habit. In practice: the figures are on the table before the decision, and everyone is looking at the same ones.
Dimension #: The angle you split a figure by: by customer, by month, by region. Dimensions are the words after “by”.
KPI Key performance indicator #: A figure with a target beside it, so you can see whether you are meeting your goal. Revenue is a number. Revenue against your monthly target is a KPI.
Measure Metric #: The number you measure: revenue, order count, lead time. Measures are what you add up or average.
Report #: A worked-out overview of one period or subject, made to be read. A dashboard shows how things are going; a report explains why.
Self-service BI #: Staff build their own analyses inside a model specialists prepared. You choose the questions; the definitions are already set.
Single source of truth #: One definition and one source per concept, so two departments do not arrive at the same meeting with different revenue figures. A matter of agreements, not technology.
Steering information #: Information that does not just show you what happened, but helps you steer better, faster or sharper. A dashboard that only looks back is reporting; only once it changes a decision does it become steering information.

Data and storage

11 terms

Data lake #: A store that keeps data in its original form, including documents and log files. Useful when you do not yet know what you want from it, but without order it becomes a swamp.
Data mart #: A bounded slice of a data warehouse for one department, sales only for instance. Users find what they need faster.
Data source #: Any system your data comes from: ERP, CRM, accounting, webshop or a spreadsheet. The more sources, the more complex and costly a BI project gets.
Data warehouse #: One central place where data from several systems comes together, cleaned and defined one way. Rule of thumb: not needed with one or two sources, worth it from three.
Database #: A store that holds data in tables. Every business system runs on one, but it is built to make the program work, not to report on.
Lakehouse #: A data lake and a data warehouse in one: the free storage of the first with the structure of the second. This is what Microsoft is betting on with Fabric.
Master data #: The core records your whole company uses: customers, products, suppliers. If the same customer is named differently in three systems, your reports will not reconcile.
Medallion architecture Bronze, silver, gold #: Three layers in which data gets steadily cleaner: bronze keeps everything as it arrived, silver removes the errors, gold is ready for reporting. Because bronze stays untouched, you can always go back.
Metadata #: Data about data: where a field comes from, when it was updated and who manages it. This is what makes a dashboard explainable.
Structured and unstructured data #: Structured data fits neatly into tables with fixed columns, such as an order line or a customer number. Unstructured data has no such shape, like an email, a photo or a pdf, and is harder to report on.
Unlocking data #: Making data in a source system accessible for reporting, for instance through a connector, export or API. Without this step, a system can hold plenty of useful figures without a single dashboard ever using them.

Preparing data

10 terms

Data cleaning #: Removing duplicates, correcting errors and filling gaps. This does not have to be finished before you start: clean what the questions you are asking actually need.
Data integration #: Bringing data from different systems into one whole. The technology is the easy part; the hard part is two systems using the same word differently.
Data pipeline #: The automatic route data travels from source to dashboard. If your dashboard is right in the morning without anyone touching it, that is the pipeline.
Data transformation #: Reshaping data into something usable: splitting columns, aligning dates, merging tables. Usually the largest part of the work, and the part nobody sees.
ELT Extract, load, transform #: The same steps as ETL in a different order: load everything first, reshape afterwards. Because the raw data is kept, you can change a step later without going back to the source.
ETL Extract, transform, load #: Pull data out, reshape it, and only then write it away. The classic order: what arrives is already clean.
Incremental refresh #: Fetching only new and changed rows on refresh instead of everything again. With a lot of data that saves hours.
Power Query #: The part of Power BI and Excel where you set out how raw data gets cleaned. You record the steps once; after that it happens on every refresh by itself.
Query #: A question put to a data source in a language the system understands, usually SQL. Power Query in Power BI translates this into clicks, but underneath it is still a query running.
Refresh #: Fetching new data so your dashboard is current. You set how often it happens: daily, hourly or near live.

Modelling

12 terms

Aggregation Aggregating #: Combining figures into a higher level: adding daily revenue up into monthly revenue, counting orders into a total per customer. Every dashboard does this behind the scenes, turning individual rows into the overview you see.
Calculated column #: A column worked out once per refresh and stored per row. Handy for filtering, but it costs memory: for adding up, a measure is better.
Data model #: How your tables hang together, plus the calculations on top. On a poor model a dashboard never becomes reliable, however good it looks.
DAX Data Analysis Expressions #: Power BI’s calculation language, where you define your sums. It looks like Excel formulas, but works on whole tables at once.
Dimension table #: The table of descriptions you split those events by: customers, products, dates. Short, and rarely changes.
Fact table #: The table with everything that happens: every order, every booked hour, every sales line. Long, and it keeps growing.
Granularity #: The finest level you store data at: per order, per day or per month. You can always roll up, but never break down finer than what you stored.
Measure #: A calculation that only runs when you look, and moves with your filters. So one “revenue” calculation gives the right answer per month and per customer by itself.
Relationship #: The link between two tables. Click a customer and the revenue chart changes with it: that is a relationship at work.
Semantic model #: The layer between your raw data and your reports that holds what everything means. Because “revenue” is defined here once, it means the same everywhere.
Snowflake schema #: A star schema where the surrounding tables are themselves split up again. It saves space but makes the model slower and harder to read. In Power BI you usually pick the star.
Star schema #: The standard shape of a data model: one table of events in the middle, tables of descriptions around it. Power BI is built for this, and it is usually the difference between a fast model and a slow one.

Microsoft platform

15 terms

Azure #: Microsoft’s cloud platform for storage, databases and computing power. For BI it becomes relevant once you need more than Power BI can handle alone.
Data Factory #: Microsoft’s service within Azure and Fabric for automatically moving and transforming data between systems, without writing code yourself. Relevant once you want to bring several sources together on a fixed schedule.
Gateway #: A bridge between Power BI in the cloud and systems running in your own office. Without one, the cloud cannot reach your own server.
Import, DirectQuery and Composite #: The three ways Power BI handles your data. Import copies it and is fastest, DirectQuery queries the source live so it is fresher but slower, and Composite mixes both. For SMEs, Import is nearly always right.
Microsoft Fabric #: Microsoft’s platform putting storage, processing and reporting in one place, with Power BI as the reporting layer. The direction Microsoft is building in, though not everyone needs it today.
OneLake #: The central storage in Microsoft Fabric, holding all your organisation’s data in one place. The idea: store it once instead of a copy per application.
Power Automate #: Microsoft’s tool for automating actions between systems, such as sending an alert when a KPI drops below target. It makes a dashboard something that warns you rather than something you have to check.
Power BI #: Microsoft’s BI platform: connect data sources, model them and publish dashboards. It comes as Desktop for building and the Service for sharing.
Power BI Desktop #: The free program where you build dashboards. Building costs nothing; you only need licences to share.
Power BI Premium #: A heavier licence type that buys capacity for a whole organisation instead of per user, with larger models and faster refreshes. Only relevant from dozens of users or large datasets upward.
Power BI Pro #: The licence that lets you share reports and use the Power BI Service, per user per month. For most SME teams this is the licence you need.
Power BI Service #: The online environment where published reports live and where you manage refreshes and permissions. This is where colleagues look.
Power Platform #: Microsoft’s umbrella name for Power BI, Power Automate, Power Apps and Copilot Studio: four tools that share the same data and the same security. For most SME projects, Power BI and Power Automate are the two that actually matter.
Qlik #: A BI platform alongside Power BI, known for its associative data model that lets you click through from any figure without predefined paths. Optilise builds on Power BI, but the underlying idea, connecting data instead of stacking it, is the same.
Workspace #: A walled-off area in Power BI where a team manages its reports. It keeps departments apart, and finished work apart from work in progress.

Roles and job titles

5 terms

BI developer #: The person who actually builds dashboards and reports in Power BI: setting up data models, writing DAX, designing visuals. The link between a data model and a screen a manager understands.
Data analyst #: The person who examines data to answer a specific question and explains the outcome to whoever decides on it. Unlike a data engineer, a data analyst does not build systems, but uses them.
Data architect #: The person who designs the overall shape of a data environment: which systems talk to each other, where data is stored and in what structure. On a small BI project this role often overlaps with the data engineer.
Data engineer #: The person who builds and maintains data pipelines and data warehouses, so data is reliable and arrives in the right place on time. Without a data engineer, every report has to be updated by hand.
Data scientist #: The person who builds predictive models with statistics and machine learning, such as expected demand or the risk a customer leaves. Only worthwhile once your data foundation is already in order.

Governance and security

8 terms

Data governance #: The agreements on who manages which data, who may see it and what terms mean. Without them, dashboards drift apart on their own.
Data lineage #: The traceable route of a figure: which source it came from, what steps it went through along the way, and where it lands in the report. Essential the moment someone asks why a number differs from last month.
Data management #: The set of agreements and processes for collecting, storing, securing and keeping data accessible. Good data management is the precondition for reliable dashboards, not a separate step alongside them.
Data quality #: How complete, correct and current your data is. One figure that turns out to be wrong, and the dashboard stops being opened.
EU Data Boundary #: Microsoft’s commitment that business data from European customers stays in Europe, in force since 2023. The answer to the question every BI project raises: where does our data actually live?
GDPR AVG #: European privacy law for personal data. For BI it means: collect no more than you need, record why you keep it, and make sure not everyone can reach identifiable data.
Row-level security RLS #: A setting that decides which rows of data each user sees. Every account manager opens the same dashboard and sees only their own customers.
Sensitivity label #: A marking showing how confidential a report is, which travels with it when someone exports it. A confidential report stays protected even in Excel.

Analysing and using

6 terms

Augmented analytics #: Analysis where AI surfaces notable patterns and explanations by itself. In Power BI you see it in features that tell you unprompted which factor mattered most.
Data visualisation #: Turning figures into charts, maps and tables that can be understood at a glance. Good data visualisation picks the shape that fits the question: a line for a trend, a bar for a comparison.
Drilldown Drillthrough #: Clicking from a total into the detail beneath: year to month, region to customer. So you see not just that a figure is off, but why.
Predictive analytics #: Working out from your history what is likely to happen, such as expected demand or the chance a customer leaves. Worthwhile once your history is sound.
Slicer Filter #: A control that lets the viewer zoom in on one year, one site or one product group. It is why one good dashboard replaces dozens of separate reports.
Waterfall chart #: A bar chart showing how a starting value increases or decreases step by step to an ending value, for example how gross margin becomes net profit. Each step gets its own bar, so you see exactly which factor contributed most.

Do you want to grow with data?

Curious about what we can do for you? We are happy to show you how data can help your organization grow.

Get in touch More about us