You Ask, I Answer: Analytics Uncovered By Data Science?

In today's episode, Monina asks, what kind of analytics can data science uncover? This is an interestingly worded question. Umalytics fundamentally is about what happened, right? It comes from the Greek word analyne, which means to to loosen up to unlock. So you have your data, and you want to unlock that data to loosen up uh that data to turn it into something usable.

Fundamentally, most analytics, at least as we experience it within marketing, are about it what happened in the data, right? You log into Google Analytics and it tells you what happened. You load uh up your social media analytics, it tells you what happened and so on and so forth. And what's missing from that, which is the whole reason any of us have jobs, is it really is just what happened. There's no explanation of why.

There's no explanation of what variables matter. Uh you just get a pile of data, right? It's like someone back the truck up and you know pour data all over your desk. And so what data science can help with is answering those next step questions, right? Your boss comes to your desk and says, Uh, you know, you you made me a dashboard with 438 variables on it.

Uh it's like, okay, what what am I supposed to pay attention to here, right? Um, you deliver a presentation, you talk to your stakeholders, you interact with your coworkers, and they're all like, We got too much data. You know, w what should we be paying attention to? Uh or the next step after that is your stakeholder, your boss, whoever's like they say, okay, so now so now what? Right?

Or or why did this happen? You know, hey, you said that the the software says uh website visits were down forty percent yesterday. Why? Why'd that happen? What are we going to do about it?

So this is where data science can help to take on some of those extra questions and start using business skills, domain expertise, technical skills, the scientific method, mathematical skills, to start to pull apart the big pile of data and that early analysis and turn it into something meaningful. Data science can especially help with understanding what's relevant, right? So you have uh and it's not even technically it's not even data science, it's more uh advanced analytics, where say you have Google Analytics. Google Analytics has hundreds of dimensions and metrics available to you. If you go to the Google Analytics website, you can actually see lists of all these things that you can analyze, and the it's kind of like Microsoft Word in a lot of ways, right?

There's so many different options because so many different people use it in so many different ways. So you have to figure out which of these variables, uh, these metrics and dimensions that are available, which ones actually matter to your company, and that's where data science helps. Four methods in particular can help turn that huge pile of data into something more manageable. Regression, which is uh figuring out essentially what numbers uh have the highest association to other numbers. So if you have a a what's called a response variable, like conversions, goal completions, revenue, you can then do a regression, uh multiple regression, of all the other stuff to figure out which ones have that association, and those may be the ones you pay attention to.

There's clustering where you can make uh different variables sort of plot it out and figure out which ones kind of clump together. There's classification for uh non-numeric variables like uh good customer, bad customer, uh the domain name of some if you're if you're yeah, analyzing your email list. Does the domain matter? Does uh a webmail address matter? And dimension reduction, where you're using mathematical techniques to to boil down uh uh into just a few couple of dimensions the the all this data that you have.

These four techniques, and and they are data science techniques, they're advanced analytics techniques, statistical in nature, can greatly assist us in figuring out what really matters, and from there we can then dig into uh you know, why did something happen? And that's where you'll see a lot of that hypothesis testing that makes up actual data science. What once you have figured out, okay, these are the variables, these are the five things that matter. Maybe it's uh unique visitors to the website or returning users or people who visited this page, whatever the case may be, you can then set up a hypothesis saying uh if we increase visits to this page by five percent, we will see a revenue increase of 10%, right? Because remember, our hypothesis has to be a provably true or false single variable statement so that you can test it.

That's where data science can help us in our analytics process to boil things down and then set up the testing conditions that we need in order to validate and build that uh that theory, that model of how we think our data works. Now, one caveat with that is that you can't create something out of thin air, right? All of the data science work that you do in your analytics is going to be derived from your initial data source. You can do a regression, you can do measures of centrality like means and medians and modes and distributions and all these things, but you're still taking that data from your initial data set from where you got your data, like Google Analytics. And that's one of the reasons why domain expertise matters so much.

Because you're never creating anything brand new. You're always working from your initial data. If your initial data has issues with it, you need to know that in order to either supplement it, maybe to fix the data using methods like imputation, you will have to do essentially some management of the data and possibly bring in more or additional or supplementary or or countervailing data to work towards that hypothesis you're trying to prove. So again, these are things that as a data scientist, you would need to know in order to be able to get the most out of your data. You need to know the business, right?

If uh if you see an anomaly in your data and you don't know that the brand manager, you know, was on vacation that week and stopped running ads for the brand, you're like, wow, something's you know, really unusual here. You need to know what's going on in the business. You need to know what the goals are of the business. Uh, and then you need to be have enough of a command of the the statistical and technical methods to figure out, okay, what approach do we need to take when when trying to solve this problem? So to summarize, analytics can't be uncovered by data science, more or less.

What we can do is find what's relevant, start to pick apart perhaps why, um, and help reduce the amount of data clutter to those things that we can then build solid testing on to build a model of how we think our marketing is working. So, really, like I said, it's an interestingly weird question. I I think I know where it was going, and there's a lot that you can unpack in your data. So start digging into some of these methods, particularly those four the top four regression, clustering, classification, dimension reduction. Those four are going to give you a lot of mileage out of your initial data and help you clarify what's really important.

Uh, if you have follow up questions, leave them in the uh comments box below. As always, uh please subscribe to the YouTube channel on the newsletter. I'll talk to you soon. Take care. Want help solving your company's data analytics and digital marketing problems?

Visit TrustInsights.ai today and let us know how we can help you.