Business Intelligence (BI) solutions and Decision Support Systems (DSS) have been around since the 1960s since the appearance of the first databases.

Every year new technology drives new tools and techniques. In 1970 appeared the basis of modern relational databases. The concept of columnar databases was already proposed by the late 1960s as a technique to quickly measure numeric values for statistics, as it was in the late 1970s Canadian census.

It was a matter of time that relational databases incorporated techniques such as indexes, columnar oriented and in-memory tables and so many other techniques that lead to the appearance of modern data warehousing, object, and No-SQL databases.

In what tools respect, in 1991 Lotus Improve presented Pivot Tables, and every year new tools and platforms appear; desktops and internet solutions, all of them evolving on speed, data manipulation capabilities, automatized analysis, and meaningful visualizations.

No matter how many tools you want to play with, techniques you want to try, or research you make for your BI initiative at work. If you don’t follow these pieces of advice, your BI initiative is already dead before it was born.

It sounds hard but it is what I’ve found and learned in practice. Hopefully, these six pieces of advice will help you propose and make real your BI initiative.

Document Yourself First, Don’t Just Go with the Hype

Lack of understanding of DSS and BI is the main error on a BI initiative. You need to draw the picture of what you want to achieve, what is the problem, before how will you address it, or think about the solution.

You may watch all those colourful videos of dashboards made with Power BI, or Tableau and get excited. Your business needs may need something different from a dashboard. So, before, answer yourself what is your need.

Do I have well established KPIs that I want to monitor? If the answer is yes then, the next question is if you already have a standardized source of data to feed the dashboard.

If the answer is “I don’t know”, you could feed a dashboard with a stable transactional database as it may be (or we spect to be) a banking platform, but you could also get more benefits from a fact table of a star schema.

A star schema will let you expand easily when you need to add a new dimension, as it would be adding a categorization to a transaction, and let this categorization grow in time without the need to add more indexes or columns to the fact table.

But what if you want to study the past and visualize the facts from different perspectives, well, then you will need a cube. A cube that is fed from a data mart, or a data warehouse; a data warehouse that is fed with ETL processes, ETL processes that apply data quality, data quality that can only be achieved with expert domain advice, and connecting different sources of data (excel o google spreadsheets, text files, emails, APIs, other databases from different vendors, among others).

ETLs stands for Extract, Transform and Load and it is the set of processes that let you read information from heterogeneous sources, transform and normalize the data applying some rules (ie, converting ‘T’ or ‘F’ from one source, ‘1’ or ‘0’ from another source, into boolean values) to achieve a model convention, and finally loading the data into a normalized source as it could be a data warehouse.

And as you may infer, there are so many options and solutions for different problems and even business domains. And you will need at least to know the basics of everything so you can propose a BI initiative and be trusted that you won’t waste time and resources.

Meaningful resources where you can start: Kimball Techniques, BI roadmap books as it is Larissa’s.

#business #business-intelligence #advice #software-development #technology #artificial intelligence