I realised last week while discussing with a friend of mine that I wrote some articles about how to do BI, what the BI needs are and some best practices on how to setup a BI infrastructures but I never wrote an article about what Business Intelligence or BI actually is. Let me correct that in this edition of the Business Intelligence Stories.
So, what is Business Intelligence?
The story all starts with data, the data we collect from our life as an individual or an organisation, the data we collect about the life of others and about the things that surround us. Data is diverse and can range from financial facts to industrial or scientific ones. But actually… why do we collect and store data in our databases, Excel™ sheets or elsewhere? Because we want to get insight into our world and the world that surrounds us. Insight is a broad topic but we can roughly categorise it in one of the following categories:
- Understanding what is
- Understanding why something is the way it is
- Predicting what will be
Understanding what is
In order to understand what is, raw data needs to be organized. For instance having 15’000 lines in an Excel™ sheet with dates, clients and invoiced amounts is nice but does not really allow to answer questions like “How much did customer C pay me each month?”, “Which are my top 10 customers in terms of revenue?”, “What is my revenue trend per country?”.
In order to gain this insight, the data needs to be reorganized and categorized. This is where Business Intelligence starts, the ability to identify dimensions (categories) and measures or facts (the things that are important to us in these categories). Building a pivot table in Excel™ is doing Business Intelligence.
When the diversity of sources of data increases (data from web traffic, data from paid campaigns, corresponding leads from Salesforce™ up to invoices for instance) the need to align sources into a single view of reality is the next step together with data validation and cleansing (Is Mr Magoo in my billing system the same person as Mr Maggot in my contract database? Does 13.12.2013 from my accounting system represent the same date like the “13th of December 13” in the comment of an invoice?). Deriving new facts from existing ones can be often needed, e.g. ranking customers by profitability requires to divide the revenue by the cost that each customer generated before ordering them.
Finally a picture being always worth a thousand words, presenting the information in a way that speaks to the audience is the final step in understanding what is. OLAP cubes and dimensional reporting systems are good at aligning sources, categorizing data and deriving additional measures as well as presenting them in a way that speaks to the end users.
Understanding why something is the way it is
It’s a great step forward in gaining insight to be able to understand what is but the next step is to be able to understand why things are the way they are in order to be able to influence reality.
Generally, the approach is to formulate hypothesis and testing these hypothesis against reality. I believe that my Social media campaign generates valuable customers. If this is really the case then a social media marketing campaign should allow me to increase visitors to my website without lowering conversion rates into paying customers. This is one hypothesis and how to validate it.
By deriving additional information from my data (the data about my campaign and the data about my leads and conversions in this example) and presenting them in new reports and visualizations I can model my hypothesis and confirm – or not – that the hypothesis is accurate which allows me to understand why something is thew way it is.
Predicting what will be
Predicting the future is a next step in hypothesis testing. I formulate the hypothesis that by doing some actions I will realize some future results. As an example, if I increase my marketing spend on paid search then I will get more paying customers.
By modeling these assumptions and applying them to the data I gain progressively additional insight about the dynamics of the system I’m analyzing, for instance I discover that beyond a certain point I’m degrading conversion rates. By formulating and validating what-if scenarios I gain insight into the future and start predicting what will be which allows me to do better decision making in the present.