You Ask, I Answer: Proving the ROI of Analytics and Data Science?

In today's episode, Gautam asks, what is your recommendation for proving the ROI of analytics and marketing data science? Uh so proving the ROI of analytics of data science of anything data related is going to be one of the thorniest problems you run into in marketing data science. And the reason for this is think about what ROI is. It's a very straightforward mathematical formula. It is earned minus spent, the essentially sort of the net earnings, right?

Uh, divided by what you spent. That's ROI. Now, here's why this is tricky. Analytics, data science, big data, whatever you want to call it, is one step removed from the things that actually generate profit, the things that actually generate revenue, right? When you use Google Analytics, Google Analytics never ever generates you money.

Your website generates you money, and Google Analytics measures that. Well, they're measurements of tools. So they're they they are technically tools, but they are measurements of tools. It's like the frying pan is what you cook with, and the thermometer that you use, a little laser thermometer, measures the frying pan, right? But the laser thermometer doesn't cook the food.

The frying pan's the one cooking the food. So it's very difficult to understand the value of the thermometer as a discrete dollar amount. You understand the value intuitively, right? If you if you're trying to get the frying pan to a certain temperature and you don't want to burn the oil, you zap the pan and it says it's you know 380 degrees. Great.

Let's go ahead and start cooking. Or it's 500 degrees. Too hot, turn down the heat, otherwise we will burn the oil and you know make a terrible tasting dish. What's the ROI of that thermometer? Well, it's not ruining the dish, or it's making the dish better.

But how do you quantify that, right? Because again, it's one order removed. So in a situation like this, you are measuring the measurement of the actions you've taken based on data. And that's essentially the definition of how you measure the ROI of analytics and data science. So the spent part is easy.

You know what you spend on analytics in terms of hard dollars. You know what you spend on analytics in terms of labor and soft dollars. That's your spent part. That's pretty straightforward. The earned part is the hard part.

So there's three different ways to tackle the earned part. Uh and they're in decreasing order of usefulness, but also technically decreasing order of complexity. The most accurate and the most complex um way to to measure the ROI is when you make a decision, like oh, should we increase our decrease our our Facebook spend? Should we run ads on YouTube? You take you you create one version and A B test saying, okay, let's use data to inform this decision.

And the B test is let's just wing it, right? Let's just do what we've always done. And you carefully document and you run these campaigns in parallel. One campaign with uh the best analytics you've got, the other campaign with just kind of winging it and hoping it all works out. And you measure the difference, and you do this for many, many different campaigns.

Now, this requires extremely good uh data governance and marketing governance to be able to carefully document both scenarios and keep them separate and distinct, you know, laboratory conditions uh to the best of the abilities you can. And you do this for uh all kind in any kind of campaign where you have the ability to do A B testing. For some channels, this is impossible to do, right? You cannot really, you know, for organic search as a whole, for SEO as a whole, you cannot just sort of well, I guess I I guess you could half optimize a website, but that's probably detrimental to the overall nature of the company, right? The second way is to A B test different groups and teams.

And this works if you have a larger company and larger teams. You have one team that uses the best analytics practices and the other team that doesn't, and at most large companies, you can easily find a team that does does not use data to make decisions. And over time you measure each team's performance. You know, uh if you have two e commerce teams, maybe for slightly different business lines, one uses analytics, one doesn't, you can relatively easily differentiate the two. Smaller companies, that's not going to really work out very well.

The third and least good way to measure that impact of analytics is period over-period comparison. So if you are, say, a company that is not especially data-driven, uh, and you go ahead and implement analytics and data science and machine learning and all these fancy things, you make an annotation in your analytics, you know, Google Analytics, your CRM, whatever, saying on this date, we did this, on this date we did this, and you and you start to really note when you've become a data-driven company, where you've made that commitment, and then you start doing period-over-period comparisons. On you know, July of 2020, we were a data-driven marketing team, July of 2019, we were not. How does how do things compare? The reason this is the least good method is because there is a tremendous amount of potential interference.

Uh market changes, if you're running on campaigns, advertising changes. Uh certainly we all know Facebook's algorithm could have changed on you dramatically then, and so it's very difficult to come up with a clear case that analytics was the reason for all of the changes that happen when there's so much interference. So, again, the A-B testing of decisions with careful documentation is the best choice. Uh, you can do the teams if you have the capability, and then period over-period comparison. If you just have nothing else, it's not going to be great.

Um, so that's the the methodology for proving the ROI of analytics and data science. The the funny part is you're actually using data science to prove the value of data science, right? This is scientific experimentation, a hypothesis that analytics has an ROI, and then creating the testing conditions to do that. So uh good question, complicated question. It can be a very frustrating question.

So under st remember that too. It's gonna be it's gonna be frustrating. As always, please leave your comments in the comments box below. Subscribe to the YouTube channel and 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.