Hands up if you’ve not contributed in a meeting where graphs are projected on how various KPIs are progressing for the company: sales, turnover, production, quality, etc. A lot of you will undoubtedly relate. Well, we have the answer, the first part of which is called Descriptive Analytics. Descriptive Analytics: tools that provide you with a description of how business processes or various departments are doing/were doing.
This kind of tool can be used to answer questions like: Which product have we sold the most of recently? Which region has the highest turnover? Which of my products is the most profitable?
These tools use business data on traditional relational databases, data warehouses, OLAP cubes, and Business Intelligence tools, allowing you to see data from a number of different perspectives.
I would say that having this information to support company decisions would be really useful.
Is there more that can be done in an increasingly technological and fast-paced world? How many other types of analysis are possible and which sources do they use? To answer these questions, I’ve prepared an infographic (also a buzzword) that summarizes them:
Let’s explore the next set of tools, the Predictive Analytics group.
Predictive Analytics: advanced tools that answer questions about how the future might play out through data analysis.
These tools are significantly different from the previous ones, both in terms of the processing that they apply, and the results. They answer questions like: What might the sales performance look like in the upcoming period? What turnover can I expect in the future in a particular region? What is the sales forecast for a certain product in the next quarter?
In practice, these tools are used to make a prediction or estimate a value, which is then accompanied by the probability of this occurring. These are advanced statistical tools, like those used for weather forecasting, but in this case, they’re applied to business.
How is such an extraordinary result achieved? Mainly through Machine Learning. This is something that we’ll look at in more detail later. For the time being, all we need to know is that Machine Learning learns from data: to create algorithms, lots of data is needed, as this then describes the information that you need to follow up in various aspects. Only when the required data begins to show certain characteristics in terms of volume (quantity), speed (acquired quickly), variety (mixed data), truthfulness (clean and reliable) and variability (linking meaning to context), can we call it “Big Data”. This is another topic that we’ll delve into later.
With the Predictive Analytics tools, you can get an idea of how the future might look. This gives you a reasonable advantage when it comes to competition. You might think that’s all you need, but there’s more…
Prescriptive Analytics: These are advanced tools which identify the best future scenario on the basis of the analyses carried out and give you appropriate recommendations.
Using the Predictive Analysis that we saw earlier, we are able to make predictions from individual factors. We then put this into action by combining all the different results which then allow us to create future scenarios and then select the one that brings you closer to your goals. This is what the Prescriptive Analytics tools do, providing a recommendation on how to act, both strategically and operationally.
For example, a Prescriptive Analytics system could analyse the sales performance of all products, integrate this data with market sentiment (useful and relevant data can be drawn from the internet), make future forecasts and then recommend stock modulation or the adoption of new product lines in the upcoming business period. You might think that a good manager does just this, but when there is a lot of information (too much) to manage and it changes quickly, having such a system gives you a significant competitive advantage.
Unlike a Predictive Analytics project, a project conducted with Prescriptive Analytics is more complex to implement and is domain specific (company, industry, etc.). There are therefore no pre-prepared tools, as the decisions that the system recommends are based around the specific topic and the project must be tailored.
The Prescriptive Analytics tools are based on those already described above (Prescriptive Analytics), so they still use Machine Learning and Big Data.
Prescriptive Analytics projects also include self-guiding machines, for example, which constantly analyse the evolution of future scenarios and go even further than simply recommending a manoeuvre: they execute it. But actually, it’s perhaps more accurate to put these types of tools in with the next category: Automated Analytics.
Automated Analytics: tools capable of autonomously implementing the recommended action, based on the result of the data analysis carried out.
Automated Analytics doesn’t mean decisions are made in an automatic and systemised way: instead, it acts on small decisions, implementing the appropriate action quickly.
For example, an Automated Analytics system can “guess” that a credit card transaction is fraudulent and then block it, or it can independently change the prices of products displayed on a website (dynamic pricing) to make the most of the market situation.
In the industrial sector, an Automated Analytics system can, for example, decide to change a tool on an automatic machine based on the stress parameters indicating that the machine is worn down and old in comparison to a new tool; it can analyse the noise emitted by the machines and stop the machines if a malfunction is about to occur; it can improve the cycle time of a machine by “learning” why the machine is slowing down at certain points, etc..
Generally, in the industrial sector, with the Automated Analytics system, you can carry out both new actions and those that have been performed by classic automation until now. With Automated Analytics, however, there’s a key difference: this system can both adapt and react to unforeseen situations, while the classic automation system performs the pre-set actions according to the cases foreseen in the software.
Automated Analytics tools are also mainly based on Machine Learning and Big Data.
I hope I have explained and clarified the numerous possibilities offered by data analysis tools, even if just a bit.
By this point, you’ve hopefully identified a tool that might be useful for you and your company. Hopefully you’ve understood the essential “building blocks” of the overall method: of the four categories of data analysis tools we’ve seen, the first category is still partially based on traditional databases (data in tables for example), whilst the remaining three, which are more advanced in their nature, mainly use Machine Learning and Big Data. What are they, and how can they be used? This is something we will look at in the upcoming articles. Stay Tuned!