School Data a Beeline for managers

Hi, Megamind!

Still in School we were taught Data analysts, taught how to apply machine learning methods to solve practical problems. However, almost any practical problem begins with business needs and business performance.

We are not going to say that at the dawn of big data, it was believed that the main insights and applying Analytics coming soon. It certainly is, but in our practice this occurs in a ratio of 80: 20 where 80% of all tasks for the analyst or even more is born from the business.

However, as the business generates these tasks, if it business doesn't understand data Analytics? Yes, very simple. In our company we spent some time talking to business intelligence data and now various departments bombarded us with orders coming up with new applications for these tools.

On the other hand, data and Analytics, once the prerogative of large companies are now penetrating everywhere, even in start-UPS today often reflect on what these data should do.

How to use data to personalize offers and create an individual product, how to deal with the outflow or minimize risks of non-payment, as using Analytics to choose the right location for a store, how to segment employees of the company for the selection of incentive schemes or to predict layoffs, how to effectively recommend products, how to profile customers, how to work with programmatic advertising.

All of these issues are more prominent in different lines of business along with others. For example, the company has a lot of data, for example, because it works with data from telematics devices: what data do they earn? Or how to make the company data-driven, all decisions were made based on data: where to begin?

Before, all chasing case studies: successful applications of Analytics to solve business problems. But, the fact is that every business is quite unique and what works for some may not work for others, but on the other hand, the success of any case is in the details, and these details no one will tell you and, again, from business to business as again, these details may differ materially.

Therefore, all successful applications of Analytics in your business You have to reinvent yourself. But to do this You need to know about the possibilities and limitations of this intelligence, and You, as business owners, and employees of Your departments, as most applications will be to generate that's as close to your business goals.

Thus, it is important to understand not only the applications of Analytics, but in this the analyst, as well as in the formulation of the problem. How long does it take to build your model, what data is needed, what accuracy is achievable, what is the precision needed because of the business sense?

Consider a simple example: you predict a call in a call center or fraud, or other rare event. Suppose You need to get the list of candidates for this event once a day, with calls for proactive contact with Your customers, and in the case of fraud to suppress it.

For example, Your analysts have made You a model with the probability of a false positive classification of a call or fraud 10%. This means that with a probability of 10% customer, which was not going to call the call center would be classified as gathering, a client who did not commit fraud as freder.

In this case, assume that the probability of correct classification of those who call the call center or commit fraud 87%.

At first glance, the model is good. You save a lot of money by reducing the number of calls to the call center or fraud in 87% of cases. In this case, you falsely classify those who are not going to make or commit fraud only in 10% of cases.
However, we can recall that the call to the call center per day is relative to the entire customer base all the same rather rare event, however, like fraud, in a normal situation. Let's say that these actions somehow affect 1% of all customers, which is pretty close to the truth.

Meanwhile, our error in 10% is to be applied on 99% of the client base. Let's say you have 1 million customers. Then, it turns out that you a day kontaktirajte to prevent a call to the call centre or denied service on the basis of suspicion in fraud of 1 million * 99% * 10% = 99 000 customers. But if Your base of 10 million customers? But if 100?

It turns out that such precision You don't like and you're willing to sacrifice the accurate prediction of those who do will call, so as to understate the errors of false inclusion in the forecast of those who wouldn't call. Since these two quantities are interrelated.

Consider another example. You want to have Analytics built a model of the outflow. First of all, you will need to agree on what is considered a outflow. In most cases, clients explicitly do not inform the company that they are gone, they simply cease to use the services. Accordingly, if they are not used Your services 2 weeks is the outflow? A month? But two? This must be arranged in advance, because what you define as the target variable, then Your model will predict.

And at what point the model should predict the care in the outflow? At the time when the client is already a month not to use the service? Or at the beginning of this period, and can be in advance so You have time to contact the customer and try to hold it?

These and many other details determine the success or failure of the application of data Analytics in each case.

There are more and more global questions: where in structure the company to place the division on work with Analytics whether it should be OU, or it can be scattered in different functions, what should be the organizational structure of the division, so his work was most efficient, what processes are necessary, what role.

In order to answer these and other similar questions, we did the course data analysis for managers, Data-MBA.

In this course, we talk about all major data mining tools and their application in different areas of the business on the example of concrete cases about the intricacies associated with it, about the possibilities and constraints on processes, technologies and many other things necessary for the successful use of data Analytics to solve business problems.

The first session on February 16, entry until February 12. No special preconditioning is not required, we will tell all in the classroom. You can sign up here.

Article based on information from habrahabr.ru