You Ask, I Answer: How Marketing Data Science Fixes Analytics Tools?

In today's episode, Manina asks, where do our current analytics tools lack and how could a data scientist help? Well, this is an excellent question because it relies on a bit of understanding of what analytics is and what we are and are not getting. So there's a hierarchy to analytics. There's a structure to it that indicates sort of what we want to have happen. It's called a hierarchy of analytics, unsurprisingly.

Let's actually bring it up here. So uh the lower portions of the left side are from Gartner Incorporated, and then uh the right side and the top are are additions I made. The hierarchy of analytics begins at the bottom. You have descriptive analytics answering the question of what happened. And this is exactly what every single analytics tool does today.

Go to Google Analytics, Facebook, to uh anywhere you go in in digital marketing, you're gonna get a data dump of what happened. What you don't get is the rest of the hierarchy, right? You don't get diagnostic analytics. Why did something happen? Website traffic was up 40% yesterday.

Well, why? Uh you have to go hunt for it. The tool's not going to tell you why something happened. Uh and in some cases, you may not be able to even know why. There are tons of different uh marketing data points where you need qualitative research to essentially ask ask someone, ask a customer, well, why did you do that thing?

Uh and only then will you start to answer those why did it happen? Current analytics tools can't do that. Of course, predictive analytics, they're not a hierarchy. If you have a data point, what's gonna happen next? Uh this is for things like time series forecasting, what's gonna be the the trend for our topic or our term or our uh email numbers in the next six to twelve months.

After that, you have prescriptive analytics. What should we do about it? What what's the the logical next step to take? And this is remember in data science you have those four key skill areas business skills, uh scientific skills, technical skills, and uh mathematical skills. This is where those business skills come into play, uh, where you need to have a sense of what the business is doing uh in order to be able to offer that uh prescription.

Like here's what we should do now that we know what we know, we know why, we know what's gonna happen next, here's what we should do. And finally, where uh really it's the there's almost nothing in the market uh except for custom built solutions and maybe the biggest tech companies is proactive analytics, which is uh starting to use machine learning and artificial intelligence to essentially say, okay, what can we get this tool to do for us? What can we uh take our hands off the wheel and and let you know self-driving marketing, if you will, uh happen, and there it just doesn't. There's there isn't that. A lot of vendors will say there is, a lot of vendors will promise you the moon, but realistically, there is no such thing as self-driving marketing.

Um so if we think about what we ask of our tools, what we expect of our tools, we are expecting at the very least the first three rungs on the ladder, and we're only getting the lowest rung, the what happened. And so that's where a data scientist uh and our data science team will help us fix not fix, but help us uh augment what our analytics tools are capable of doing. There's so much in the data that frankly is very difficult to parse out um simple example if i if you're doing regression analysis where you want to figure out okay I have all these different data points time on site bounce rate uh traffic from Facebook traffic from email uh things and I have an outcome like goal completions on my website well what if all of this data actually matters what has a relationship to this outcome that I could test and if you have all these different data points again analytics tools today can't do that. Even though mathematically it's actually not that hard to do. It is very computationally intensive and it requires some experience interpreting the results.

So a data scientist in that case would take all that data out of Google Analytics or wherever run it through a regression model of some kind. Everything from straight simple linear regression to gradient descent gradient boosting extreme gradient boosting uh you name it there's so many different techniques to use uh the data scientist would look at the the data, look at the the distributions and things, make a decision about uh which method to choose, maybe try a few of them, and come to a conclusion, okay here is the method we're gonna use here's the the the mathematical technique we're gonna use the code run it and say now we have this outcome we care about conversions and we have these you know five hundred data points in Google Analytics. Here are the three in combination that have a relationship to the outcome that we care about. And now we can start the science part of, okay, now let's set up a a a hypothesis that these three matter most to conversion and start running tests. Okay, if we increase uh, you know, uh time on site by 10%, we should see a 10% increase in conversion.

Good hypothesis, right? It can be provably true or false. We then go and build some tests, run some tests, optimize the site using, you know, Google Optimize or whatever, to focus on a goal of increased time on site, and then look at our conversions and see did we see a proportional increase in time on site that was matched to the proportional increase in conversions? If so, our hypothesis is true. If not, our hypothesis is false.

And that's where a data scientist can really take these analytics and turn them into meaningful results that guide our business. Now, as we've talked about in this series of questions, it's not going to be cheap and it's not going to be fast because this is science. This is experimentation, this is testing. And it's not something you can just you know snap your fingers or or buy a piece of software and do. The most important value add that a data scientist offers, in addition to just being able to do the thing, is to look at the data and tell you where it's going to go wrong, or tell you that the data is just not good enough to reach the conclusion that you want to reach.

So again, that's where uh data scientists can help fill the gaps for these these current analytics tools lack. They just can't get there yet. Will they get there someday? Maybe. It would depend on how accessible the results are to an end user, because a company like Google is not going to make you know these crazy additions to Google Analytics for the benefit of data scientists only.

They need it to benefit everybody. Uh and they have the the API and the programming interface so that a a skilled data scientist or skilled data science team can extract the data for themselves and do those computations without them needing to add and clutter up the interface for end users. It's a really good question because there's a ton that data scientists can do in these other rungs of the ladder that today's tools can't do, and that's okay. It doesn't need to be in there. Um if you have follow-up questions to this topic, please leave them in the comments box below.

Subscribe to the YouTube channel and the newsletter, and I'll talk to you soon. Take care. Want help solving your company's data analytics and digital marketing problems? Visit TrustInsights.