4 Types of Analytics.mp4

00:00:00

No matter what type of analytics that you're doing or how much data that you're using analytics, typically falls into one of four major categories

00:00:09

Today, we're going to look at those four types of analytics analytics, it's a really broad term

00:00:16

It covers a wide variety of details levels and data volumes but no matter what type of analytics are performing and what you're trying to achieve, you can usually categorize at one of four ways, you are

00:00:32

typically going to want to either describe, diagnose predict or prescribed, these four types of analytics and build off of each other

00:00:41

These four stages of analytics describing diagnosing, predicting and prescribing will feel very familiar if you've ever been to the doctor because you weren't feeling

00:00:51

Well, let's look at what that might look like you went to bed last night, feeling fine, but you woke up this morning feeling any But fine

00:01:01

Maybe you're a little achy and you have some chills you suspect

00:01:05

You might have a fever

00:01:06

So you take your temperature and find that it's a few degrees higher than normal

00:01:10

At this point, you've described what's happening and now, maybe you make the decision to go to the doctor to get checked out

00:01:19

Once they're the doctor asked you questions which go over the described process

00:01:24

Again, they might even recollect information, they'll probably take your temperature themselves to check it, and they might check your heart rate in your blood pressure at the same time

00:01:34

Now, they move on to trying to figure out what this means, based on all your symptoms, the doctor diagnoses that you most likely have the flu, you might be given a rapid flu test but that

00:01:47

often doesn't change what the treatment is at this point we reached the diagnostic stage

00:01:52

Now what happens if you don't do anything, many people feel miserable for a few days but then work

00:02:00

Just fine from the flu on their own, but maybe there is underlying health problems

00:02:05

Are other risk factors that make it more likely that you won't recover is easily the doctor talks with you and text your medical records and finds

00:02:13

There's nothing that makes you especially vulnerable

00:02:15

So at this point, they predict that with rest and sometime you'll feel better on your own at this point

00:02:24

A prediction has been made that you'll be fine without any additional medical treatment, the doctor, then send you home with instructions to rest, drink, plenty of fluids and to take an over-the-counter pain

00:02:36

Reliever, if the eggs get to be bothersome at this point, you have a prescription

00:02:42

You've been prescribed rest fluids and over-the-counter pain

00:02:48

Reliever if necessary, this is how you optimize the Improvement

00:02:52

Will you get better? Even if you don't do these things, most likely but this will help it go faster and for you to feel bad

00:03:00

As you're getting healthy again with this, we've just gone through these four stages of analytics

00:03:07

We can use this exact same process to work through a problem and stop along the way wherever we need to reach

00:03:14

Sometimes, the only thing we're trying to accomplish is to describe what's happening other times, we need to predict what's going to happen or prescribe, what will happen in the future

00:03:23

Using the analogy of visiting the doctor

00:03:25

If you woke up today and felt fine, there's really no need to go on and diagnose, predict, and prescribe you feel fine, Everything feels normal

00:03:36

So the added value of doing this additional stages is minimal and the actual work as minimal as well

00:03:42

Because we would assume that if we feel okay, that everything's going to be continuing to occur, we're going to continue to feel like we have felt unless we have some other information from a data and

00:03:55

information standpoint

00:03:56

Here's what those four types of analytics look like The most basic level we can produce regular can reports that tell us what happened within the business

00:04:06

At this point, we're really just answering what happened or maybe we've gone a step further and we produce ad hoc reports that go into more tabs on how many when, and where both of these are

00:04:18

descriptive analytics

00:04:20

They provide basic expository information what when how many, how much at the next level? We start to query and drill down into our information, further

00:04:32

We not only want to understand what happens

00:04:35

We want to know why that it happened

00:04:37

You also might create, indicators, are flags to notify you when this thing that you're studying occurs again in the future, when you do that, you're asking where do we look both of these are diagnostic analytics

00:04:52

Diagnostic analytics are really focused on why something happened and how we're going to know if it's happening again

00:04:59

Both descriptive and agnostic analytics

00:05:02

Focus on the past

00:05:04

It's all about what happened before the current point in time

00:05:08

Now, we start to look towards the future when we get into Predictive Analytics and then prescriptive analytics Predictive Analytics, help you understand what variables are a play, how they relate to each other, and any correlation

00:05:22

that might be present

00:05:24

They really start to help you understand what happens in the future assuming that we changed nothing from an analytic standpoint

00:05:32

We focus on statistical modeling and then move on to predictive modeling assuming that nothing changes

00:05:38

What does the future hold in some cases, we might want to change what the future holds in that case were moving into, prescriptive analytics, prescriptive analytics, often starts with random testing, but if we just try

00:05:52

this and then it moves onto optimization

00:05:56

What's the best course, forward? What actions show? Do we take in prescriptive analytics? We're trying to predict the future based on changes to what's happened in the past

00:06:06

I mentioned that these four types of analytics build off of each other

00:06:11

If you want to predict what's going to happen in the future, you have to start with the scribing and then diagnosing, if you jump straight to predicting your predictions, likely not going to be very accurate

00:06:24

or not, more accurate than random chance would allow for this reason, make sure you take time

00:06:29

If you're predicting to also understand in detail what happened and why it happened

00:06:35

But the only way to really understand what's going to happen, going forward

00:06:39

The same thing with prescriptive analytics

00:06:41

People often want to jump straight there because they want to make changes or they assume they know what happened but if you don't take the time to understand what really happened and why it happened and

00:06:53

predicting what happens if nothing changes, then your prescription is likely to not being Credibly effective, you're going to miss out on key factors that you should have accounted for because you didn't take time to study

00:07:07

the past correctly

00:07:09

Like, I mentioned, this doesn't always have to be a long process

00:07:12

If you woke up feeling fine this morning then you could run through the entire process and to end in a matter of a few seconds

00:07:21

You would describe that everything feels normal or like you're used to

00:07:24

You would diagnose that everything's

00:07:27

Okay

00:07:27

I'll predict that, it will continue to be fine and prescribed that you keep doing what you've been doing, even when something's wrong with this process, can go quite quickly

00:07:37

If you woke up this morning and you had a headache, you might describe that

00:07:42

You have a pressure in your head, you would diagnose

00:07:45

That is a headache

00:07:46

You would predict that overall you don't feel too bad

00:07:50

And so you're probably be fine throughout the day and then you'll prescribe that you either tolerate it without doing anything or you

00:07:57

Take some specific actions that You know, if help to mitigate headaches in the past, all of that could probably happen in 5 to 10 seconds at most

00:08:06

Usually, any analytics activity will be pretty clear from the start

00:08:11

What's the age that you need to read? You'll know whether you just need to describe whether you need to diagnose and explain

00:08:18

The why, or whether you need to predict, or prescribe what to do in the future that we've covered descriptive analytics by Gnostic analytics Predictive Analytics and prescriptive Analytics