How to implement a BI-approach to data analysis: practical recommendations

The article was prepared by Sergey Shopik, director and founder of the Client Experience Laboratory. Based on material published by Martha Bennett on the website of the international analytical agency Forrester Research. We invite everyone to June 18 at 20:00 Moscow time. to the free master class "The process of visualization from A to Z". Sign up here .

Too little data. Too much data. Incomplete data or limited access to them, reports and dashboards that have been formed for too long, which often do not meet the goals set. Analytics tools that only a few trained specialists can use. All this - a list of complaints from the field of data mining and business intelligence (BI). It is extremely long, and automation, unfortunately, does not serve as a solution to these problems. At the same time, BI has been one of the top priorities for several years for implementation in organizations, as companies are beginning to clearly recognize the value of data and analytics when it comes to optimizing solutions for better results.

So, what can you do to ensure that your BI initiative does not end up in a failed project dump? Finding the answer to this question is not something unusual and difficult, but it will require answers to clear questions and the separation of the “chaff” seeds. It is often enough to hear stories of how multi-million projects in this area have suffered a complete fiasco. Often this was one of the following reasons that we will try to figure out.

What is the difference between a successful project for implementing BI analytics and a project stuck in a production hell? Studying the best practices of successful projects, the difference may seem obvious, but it is the differences that distinguish those whose BI projects do not meet the needs of the business (or fail in principle) from those whose projects succeed.

And so the most important thing: to which category of tasks will we assign a similar project? To corporate IT or to one of the business units whose reports we want to automate and whose data we want to look at? Usually the whole problem lies in the fact that the implementation of the project is completely at the mercy of corporate IT without involving business users in the process. Moreover, this happens often at the initiative of the latter - let them implement it, and then we will press one button and the “analysis” will begin. Not really. The initiative should go precisely from business and business objectives, but not the other way round. Obvious, but difficult thing. How do we do this?

• Form clear tasks: for WHAT to me this dashboard, with what PURPOSE will we consider this or that indicator? The bad answer is to appease shareholders / management. ” A good answer is to evaluate the effectiveness of certain actions and on the basis of this, make a, b, c decision.
• Be flexible and do not try to close all tasks "at once". The second common mistake is writing the perfect TK. Automate one task, check the result and go to the next. Do not try to deploy a large-scale project for a year. Reach the goal in 12 steps - one step per month. Do not forget about the main goals and objectives.
• Understand the data. Projects do not take off due to lack of data. And this also happens. I remember a case when the client base segmentation project stopped due to the fact that in the company's accounting program the sales data was not tied to discount cards! Does this mean that you need to sprinkle ashes on your head and abandon your plans? In no case! This means only that it is necessary to link the data and begin to accumulate it, in order to later approach the intended one. At the same time, start building a system of indicators, which you will later integrate into the processes. Check if all the data is there and repeat the cycle.
• Select tools based on the task, not vice versa. Old joke: we bought something, and now we are trying to push our processes there. It should be the other way around. The tool for BI-analytics is selected for the task, and not vice versa. Well at the same time, if all units in the company will know about the technologies that are used in it. Otherwise, you will have 20 programs in each department, each one considers his own way, but there is no single version of the truth. All you can do is implement the 21st tool.
• If necessary, do not be afraid to resort to external assistance. Especially nothing to add. Normally: call a consultant or bring in an external team to resolve a necessary issue in a short time. Not normal: do everything yourself and do it for two years.
• Change management and learning is a continuous process. You can not implement something "for centuries". The market is changing, changing goals, indicators, the situation in the company. It is important to monitor the relevance of decisions and develop them promptly when necessary. In the end, what works for a separate shop is unlikely to be completely relevant for a large federal network.

Where best practices are present, by definition, there are pitfalls that should be avoided. We identified the most common among them:

• Using IT in business intelligence seems simpler than it actually is. Until you have implemented a BI approach to data analysis, you usually have 1 question. After implementation, the number of questions increases tenfold, because opportunities for analysis becomes many times greater. Together with technology, develop a culture of working with data.
• Refusal to perform their duties after attracting external partners to assist. So, unfortunately, it does not work. An external consultant or team will help you build a process. But the responsibility for him and work with him is for you.
• Focusing on technology development and implementation, not on change management and training. It is better to implement a small reporting system for Power BI and efficiently make operational decisions based on data than spend hundreds or thousands of hours implementing SAP and not use its functionality even by 1%, but continue to send tablets to Excel in each other.

Follow these simple rules and you will no doubt succeed. And the transition from decision making on a whim, to weighted and digitized decisions using BI-analytics will be as painless as possible. Successes, friends!