Customer Engagement

How to analyze big data? 5 easy steps.

When in 2013 Netflix shared their House of Cards investment details, we were all – well, let’s just say – alarmed. $100 million, Kevin Spacey, and David Fincher put together to waste on a TV series? Oh, come on! As it turns out, that made more sense than some of us may have thought back then. Why?

It’s no accident that this particular show changed the course of TV industry. They did their homework and analyzed data properly. Even though the latest season didn’t get such a warm welcome, it was a game changer for Netflix. The brand’s name even became a proprietary eponym (as in #netflixandchill) which means it has become a commonly used generic name for all similar platforms. In the marketing world, that’s huge.

Finding what truly matters

The difference between pre-Netflix and post-Netflix times lies in the audience’s needs. They became the ones who have finally addressed the current needs of Millennials. And they did it right. However, they didn’t do it the old-fashioned way – asking a bunch of teenagers about some likes and dislikes. What they do on a daily basis, is collecting and analyzing quite an impressive amount of website events data, such as how many people saw or paused and rewind each show, what else they watched, and when. They gather data to draw conclusions about the likelihood of the success of upcoming movies and shows. And their data-driven approach pays off. The House of Cards’ $100 million dollar investment was earned back in about three months and apparently, they have better overall ROI on their shows than any other entertainment studio, as only 30% of their original shows get canceled after the first season.

2D data growth

According to commonly quoted IDC report, by the year 2020, there will be 44 zettabytes of data generated all over the world. And it grows exponentially as the ability to manipulate data becomes more and more mobile. Not to use opportunities that come along, wouldn’t be wise when almost every CEO has the words big data written on their whiteboards with a permanent marker. The trick is to use it skillfully.

From what I gather, the development of the internet and new technologies goes in two dimensions. One being the – mentioned above – extreme quantities of uploaded data. The other comes as a natural consequence. It’s the need for quality. We no longer get excited about the amount of Facebook fans. We want them to be engaged and loyal. Even outdoor marketing, which was always known to aim at the widest target, is starting to localize.

Since we’ve passed the big data hype and are now facing the sceptical phase, what we need to do, is stay focused on how to make the most of this trend. Not to stay behind but don’t fall for it blindly either.

How to approach big data to gain truly relevant insights?

The best way to put these oceans of data to good use in business is the analytics. Some companies go as far as to hire proper data analysts. Which apparently is not a piece of cake, even though you can easily find big data analysis courses online. However, taking a few online classes is not enough to become a professional. Even if you don’t plan to expand on such a large scale, what you do need for sure – according to Bob Gourley, the author of Data Divination: Big Data Strategies – is a strategy. Having gone to the end of the internets for tips and tricks, I dare to propose these five easy steps in handling big data.

1. Divide up

Custom audiences have become a very hot topic recently. You need to personalize email campaigns, up-sell, cross-sell offers. Your imaginary friend – a Buyer Persona – has come to your party with family and friends. The key to personalization of your communication is acknowledging that amongst many people that you want to reach, each of them is different and has different needs. Sure, it’s not possible to personalize 1 on 1 but segmentation of your target to small groups might be just fine conversion-wise. The more data you get, the more evidence to cluster. So approaching big data, don’t be scared. Think of it as a huge pile of pretty small bits that give you a wide variety of reinforcements.

2. Spread out

Since you already know you want all kinds of target groups, you might simply jump into analyzing these diverse data sets. You have plenty techniques to choose from, depending on your business goals and whether you have structured or unstructured data to deal with. It’s always good to double-check, though. So, you can mix and match your ways to find relevant insights amongst your data. When it comes to business intelligence, you might want to check:

  • Data mining – is a way to find new patterns in data; it’s based on assumption that if something is repeatedly occurring, it might be significant somehow.
  • Cluster analysis – this may be the next step since it’s a way to group objects determined by certain sets of similar attributes.
  • Predictive modeling – this is what a psychic would do but much more data-driven. We’d probably all agree that there’s a slight but quite significant difference between a fairy and a weather forecaster. With this approach to big data, you just take a chance on the theory of probability and the odds that it will deliver are, well, pretty big.
  • Textual analysis – a skillfully crafted natural language processing algorithm can extract very useful data not only from digits but also from multiple blocks of text; needless to say, the sentiment of online mentions about – let’s say – your brand, or gender of the authors, or their locations might come in handy.

3. Catch up

Act in real time. It’s no secret that instant updates are crucial in successful business. Even if this step might seem indistinctive for big data, it’s not so obvious that with huge chunks of information your analysis is going to be flexible enough. You can spot otherwise completely fine analytics tools which sadly provide updates that you have to wait hours for. However, for example in e-commerce, it’s a common practice to base on big data when creating a dynamic pricing. If you feel like experimenting, try to book a flight on Friday and then go and see the same offer on Monday or Tuesday. That just shows that real-time data analysis is available. And it’s worth every penny.

4. Suit up

Well, to be precise, your data should suit up. As in dress nicely into eye-catching charts and graphs so that you would no longer waste your time trying to draw some conclusions. Especially if you’re dealing with enormous amounts of digits or online mentions. What you need to do is to find a proper analytics tool that can provide you with comprehensive visualisations of your data. This way you can understand it easily and act on it. Take for instance critical alerts in online listening. With SentiOne you can set up a number of negative mentions to trigger the alert. Time saved. Easy.

5. Watch out

Even though you can save some time and money thanks to big data analysis, you need to keep your eyes open. There are a few issues that come along with meddling in what people are putting on the internet. First, there’s the privacy matter. The whole IT world is walking on tiptoes around it. However, as long as you use a genuine platform to gather and analyze the data, you are protected. What you do need to remember, though, are typical statistic mistakes. Like for example causation-correlation puzzle. Even if you use a proper tool, make sure you set it up correctly. Did you know that Watson of IBM, among many uses, was a foundation for medical diagnosis tool? It would gather information and narrow it down to the most likely outcome. Nevertheless, you cannot say the program is a full-fledged physician.

Speaking of health, what you do need for sure is a healthy big data policy. A strategy that will help you tame your big data and use it to your benefit. If you’ve heard that big data is a magic device that will skyrocket your business, you may want to rethink your source. For every excellent background research group comes a Dr. House that has to connect the dots. Because with handling big data, a human touch is needed more than ever. Sure, deep learning is fast and efficient but at the end of the day, it’s a real human that makes the decision. The more data-driven, the better.