“In God we trust, all others (must) bring Data”. I believe it was William Edwards Deming, a statistician ahead of his time, who said this phrase that says so much in so few words. Today their validity is, if possible, much greater than in their time. We are undoubtedly in the information age, there are many who say that data is the new oil.
In any case, like oil, raw data has little value, and can even be extremely annoying, dull, and even harmful. Hence, science and engineering and even the world of economics have been working for decades to obtain the formula to clean, process and extract information from that data, to establish the necessary processes to make that information useful.
Clean data is always the starting point. Sometimes we ask ourselves, but is there dirty data? Data, like everything in our world, gets dirty due to the mere passage of time, dust falls on it, we leave tangles and junk in the way, for example, a customer signs up who was going to buy from us but never sent your authorization. An order is managed, but there is no stock, it remains in limbo, the client no longer wants it, no one cancels it, we gave the client CIF 1111111-A because the next day he was going to send us the information and it never arrived. All of this is part of everyday life and is dirt, which must be eliminated, or at least treated so that the rest of the information shines in all its splendor.
From this information we will be able to do various types of analytics.
Useful information
Descriptive Analysis.
The one who tells us what happened, adds and summarizes the information, graphs it, and shows it so that we understand and process it. It is basic analytics, the pillars of analysis. With a deep knowledge of the domain (the business environment in which we are operating) we can also begin to draw conclusions. We will know that there are products that are more successful in some places and are not distributed at all in others, we will be able, in short, to take corrective measures based on data.
Predictive Analysis.
A predictive analysis is one that shows us a most probable future scenario, one that tells us what is going to happen. In the distribution sector, for example, it could give us a consumption forecast to properly manage the supply chain. Not that the 100% of sales will be right, but it will obviously be better than no forecast at all. If for this prediction we feed the data with precise and clean information, the predictions will be much better and closer to reality.
Prescriptive Analysis.
A prescriptive analysis is one that tells the recipes to apply to achieve an objective. It is the analysis of levers in which we can touch certain levers to achieve an objective. An example could be avoiding customer abandonment. Let's imagine that we collect a series of variables about the customers who leave us, let's imagine that these variables are for the last 6 months, number of complaints, average delivery time from order, number of purchases, trend (up, down or neutral), data of the competition (we are more expensive or cheaper, etc.) and some more that we determine are relevant. From there we could obtain the probability of a customer abandoning us based on the data of customers who have already abandoned us. With this we would have the first two analyses, the descriptive one, with the data of the variables we were talking about and the predictive one, we would know if the client left or stayed with a probability. Now we can establish loyalty campaigns, discounts, bonuses, service improvements, whatever we determine and measure the effectiveness of those measures. He analysis It would tell us that, if you change this or that variable, the client will no longer abandon us.
Diagnostic analysis.
On other occasions it is necessary to do an analysis of causes. Why did this client leave? What did we do wrong to make him abandon us? This type of analysis is based on the detail of the data, on the differences between some subjects and others and will be very useful to put measures that improve situations that we do not want.
But... isn't this something for computer scientists and statisticians?
Well, the truth is that we no longer, or at least, we do not have to put ourselves in the hands of third parties to carry out a good part of the above analyses. For some years now, software manufacturers have been very concerned with what they call "democratization of business intelligence." This democratization is nothing more than the creation of tools with which mere mortals, far from the Olympus of mathematicians and engineers, can , we can carry out each and every one of these steps, we will be able to obtain the information from the data source in which it is, from open data sources, government, climatological, statistics institute, to companies' own databases, Excel sheets or files exported from the ERP.
Once this information is obtained, we can clean it, transform it and enrich it, that is, we can remove clients and orders that never came to be, we can summarize it by month and we can exploit it in graphs and tables. Surely many of us are thinking about Excel, the most used copy-paste tool in the world, yes, but also others that can solve this problem much more effectively with substantially larger volumes of data. In my case, PowerBI has been a real revolution in what has been called SelfService BI.
Contained in this tool we will be able to do, at least in a first step, a good part of the four types of analytics that we have just explained. We will of course be able to do descriptive analysis, as can be seen in the graph, where we are obtaining the key metrics to understand and measure our business, evolution data and comparisons between elements, in short, a lot of descriptive analytics.
We can also do predictive analysis, some of its components allow us to graph futures using time series algorithms (below they use multiple regression, but we do not have to know the mathematical foundations to use them), we can also look for anomalies, data that was not likely to occur. will happen or that go beyond what was expected.
We will be able to do prescriptive analysis, using What-if type analytics so that we can propose different scenarios and see the result if these factors occur and we will also be able to do causality analysis through some components that perform complex statistical calculations. for us.
In summary
We are in the era of data, and we have also been granted enormous power, the power of tools that bring very complex mathematical calculations to our daily lives, used from the point of view of a conventional user who knows his business and can provide a lot of value, but also with the rigor that this type of processes require.
¡Train yourself with EIP, the No. 1 school in employability!
