Big Data and analytics are hot topics today and analytics projects tend to give big promises, but how to get to real results without wasting organisations time and money and how to ensure that the project meets its full potential?

 1. Identify real business case

Real world business case sets a clear and understandable goal when starting an analytics project. It also moves the focus from the actual method and hype phrases like Big Data into actual business issues. In the end of the day that’s what it is all about, business. It is also much easier to communicate, get people motivated and to understand the process of using analytics when there is actual business case in the core of the Project.

When new initiatives and methods are introduced into the organisation their value needs to be proved and often some organisational change is required. Analytics is no exception, in fact it is sometimes very hard to get people on the analytics wagon since data analysis and predictive analytics are based on scientific methods and for that reason not so easy for everybody to understand.

Cultural change is a hard thing to do and the only way to make it is through small success cases where the business value of analytics is proved with real relevant business cases. Too often analytics projects die before they get started or fail to realize their full potential due to concentrating too much on the theory or method instead of the valid business case.

You can use various methods of brainstorming to identify a business cases. Business cases should be evaluated against their return on investment. How much time and effort experiment requires compared to the size of the possible effect on selected business metrics. When you start experimenting analytics with a real world business case you’ll realize the value of analytics much faster.

Too often analytics projects die before they get started or fail to realize their full potential due to concentrating too much on the theory or method instead of the valid business case.

2. Make all your data accessible

In many companies business people are struggling to get access to the data they need to support their decision-making. Operative applications and databases may be outsourced or delivered as software as a service model and when that is the case the data is often analyzed only at the source and not by integrating all the data from the sources to get real insight. Your data science team and decision makers need to access all the data and they need to access it when they need it.

Analytics is about probabilities, so there needs to be just enough quality data in order to build good enough models . Often the mistake I see is that too much emphasis is given to the data quality when the organisation thinks that the data has to be perfectly clean and consistent in order to get results from analytics. Sometimes getting started with the use of analytics gets postponed till the data warehouses have been built, which may cause a significant amount of money being left on the table.

In general, traditional data warehouse is just too slow and complex with all its service layers and may be a major obstacle trying to get the data in the hands of the analysts when the business requires results. Existing data warehouse is a good data source for analytics, but in modern data management platforms where a lot of analytical workloads are run, most of the data bypasses traditional data warehouse entirely due to the fact that most data already comes outside of the company from external and unstructured data sources.

3. Use a process to get from Insight to Action

In order to get results from analytics fast its important to get into actions based on the analysis. Models don’t have to be refined into perfection but instead its better to start with a good enough model and refine it later, changes in the model will happen over time anyway. Analytics is a lot about experimenting and learning and some lean principles are good to be in place right from the start.

It may sound that you just run your experiments and see what works and what doesn’t, but experimenting wildly all over the place is not the way to get to results fast, instead there needs to be a clear process how the experiments are managed and validated. The tools for managing the process doesn’t have to be complex, simple Kanban board with an experiments backlog is enough to keep the experiments in order and ensure that the validation is done and learning from the experiments are taken back to the process.

It’s inevitable that some experiments will fail but sometimes failing gives more insight than success. Failed experiments should be revisited over time, because when the experiments are run in a rapid phase some experiments may prove to be successful with minor modifications when the environment changes and more learning about current state of the business happens.

No matter what your process and tools are, the most important thing is to start making actions based on analytics today – your competitors are already doing it.


Author: Jouni Leskinen

A bit about me. I’m the director of Research and Development at Madtrix. Concentrating on growing our business as the solution for complicated, manual marketing reports and dashboards that make all our lives much more difficult than it needs to be.

Madtrix brings you easy to understand, omni-channel data integrated marketing dashboards with data platform to collect all your data for better marketing.

To speak directly to me email