Data Analytics

00:06 Data analysis is the most important part of making decisions based on data or insights.

00:13 What's the point of collecting so much information if you don't learn anything from it? In reality, this doesn't happen all that rarely.

00:21 First, all possible data is gathered, but the second step, analyzing the data, is unfortunately not done very well.

00:30 But this shouldn't be a surprise, since data analytics is a broad field.

00:34 It's not easy to live there.

00:37 Five questions can be used to organize the data analysis.

00:41 These are like steps, and as you go up them, they get harder.

00:46 What happened? That's the first question.

00:49 These so-called descriptive analytics.

00:51 The second question is why something happened.

00:54 Diagnostic analytics, in short.

00:56 Third, what does this teach us? Finding analytics.

01:01 We are already in the future, so what will happen? Analytics for the future.

01:06 And finally, fifth: What can we do to make something happen? What is known as "prescriptive analytics." So, data analysis is the process of getting answers to these kinds of questions from the data that has already been collected.

01:21 The questions in steps three through five are much harder, but they are also a lot more fun because they lead to actions.

01:29 Let's look at an example of the whole thing to see how it works.

01:32 Let's think about a hospital and what these five stages would mean for getting materials.

01:37 The future will get harder the further you look into it, of course.

01:41 First, descriptive analytics would include things like the cost of the material over the last year or the cost per patient.

01:48 Second, diagnostic would be an investigation into why costs went up in the last quarter.

01:55 Third, the lessons learned from discovery analytics.

01:59 For example, if I want to keep my suppliers, I should pay on time.

02:03 Fourth, predictive analytics.

02:06 This could be, for example, making cost predictions based on how prices have changed over the past few years.

02:12 And finally, fifth, prescriptive analytics would be: How can we cut costs? So, for example, by having fewer suppliers.

02:19 Then, different situations would be acted out here.

02:23 There is a basic difference between two types of data analytics, and the way they are done is also very different.

02:31 On one hand, these are just ad-hoc analyses.

02:34 These are unique analyses.

02:36 And repeated analyses, also known as "routine reporting." The KPIs, or key performance indicators, are an important part of routine reporting, which we will talk about again later. First, we'll look at the ad hoc analysis.

02:51 It's about things that are happening now.

02:54 For example: What are the reasons for last month's drop in sales? So, first, you get the data you need from the sources, ideally a data warehouse.

03:04 Then, the analyst tries to find the causes by using the right methods, which can range from simple compilations to complex statistical methods like clustering.

03:14 Most of the time, the result is a presentation or an Excel spreadsheet.

03:20 Routine reporting, on the other hand, answers questions that come up again and again. How, for example, can the current data be summed up on a regular basis? This can look a lot of different ways.

03:33 It could be just a table with the most important numbers, or it could be a PDF file with charts, tables, and dashboards.

03:41 The dashboards, on the other hand, can be anything from simple, static collections to interactive graphics with filtering options to self-service environments where trained users can put together the corresponding graphics themselves.

03:56 All of them get regular updates.

03:59 This should, of course, be done automatically.

04:02 If you don't, you'll waste a lot of time and work resources.

04:05 But this makes them less flexible, and you can't answer each question as individually as you can with ad hoc analysis.

04:13 For routine reporting, you need a set of key figures.

04:17 These key performance indicators (KPIs) should shed light on the reporting area from different angles. Creating good KPIs is an art in and of itself that comes with some risks. If you evaluate a department based on key performance indicators (KPIs), then these KPIs have a controlling effect.

04:40 So, it could be that the focus is on getting the key numbers right.

04:44 Take the retail industry as an example.

04:46 If we were the only KPI that had sales numbers compared to before, Category Managers would probably take a lot of actions with especially strong discounts.

04:55 This leads to more sales, but it costs the dealer money because the margin goes down to minus campaigns.

05:03 Retail is a good place to see this so-called "action addiction." KPIs give a strong sense of control. This is good at first, but it also needs to be well-balanced so that it doesn't just optimize one side.

05:18 KPIs can be very different from one another.

05:21 The department head's job is then to move and gauge within this framework.

05:26 One should not forget that the company accepts KPIs.

05:29 This can be ruined, for example, by making too many demands.

05:33 So, what good are KPIs if the target values can't be met, the calculations aren't understood, or the data on which they are based is questioned? So, it's important to sit down with the department to talk and clear up any questions or concerns. In short, data analytics is the process of looking at existing data to answer certain questions.

05:57 From simple summaries to predictions and simulations, you can divide these questions into five steps that get progressively harder.

06:07 Ad hoc analyses are ones that are done just once.

06:11 Routine reporting is used to analyze and show key numbers on a regular basis.

06:17 To get the right tax effect and avoid one-sided optimizations, it is important to find the right balance between these key numbers.

06:26 Data science includes data analytics.

06:29 This explains the whole process, from finding the right question to building the right data pipelines to automating.