Hey everybody in this article we want to run you through the first steps of applying Machine Learning and Artificial Intelligence for business and growth. We’re gonna look at what are the “must-haves” of AI and machine learning for business and growth, What are the “should haves” and what are the “nice to haves”. So at growth tribe we train people in their first steps with AI for marketing and growth. Let’s use a small analogy.
Now just like the Industrial Revolution took us from one horse to 450 horses, AI and Machine Learning are taking us from one brain to thousands of brain working simultaneously to help us answer business questions. The first step to using AI or machine learning for business is typically to know what questions you need to answer.
- “How likely is that person on our website to buy our product or service?”
- “How much is this customer likely to buy this year?”
- “Which one of our customers is going to stop using the product soon?”
- “What characteristics should I segment my customers by?”
- or even more recently
- “What are the main personality traits of my customers?”
Now although AI and machine learning are like a thousand brains helping you inside your company to answer these questions In real time. All of these different questions are actually more or less mature in the marketing and growth sphere. Some of them are really mature they’ve been used a lot there’s many use cases, some of them are less mature they’re up-and-coming and there’s actually not that many use cases.
So we’ve actually gone ahead and mapped out for you how mature each of these applications is for marketing and growth. Just to make it visual and simple. Now at the very top we have predictive analytics. Predicting outcomes – often future outcomes – based on historical data. Predictive analytics allows Marketers and Heads of Growth to predict the customers life time value, to identify customers that are more likely to be loyal and, of course, to predict whether a lead is a good value or not – how much resources and time should I spend on each specific lead.
It’s also allowing us to predict how much a specific customer or group of customers will be worth throughout their whole customer lifetime. The reason it’s at the top of our chart is that it’s quite easy to implement and it’s been proven again and again and again. What’s also fantastic is that you don’t need that much data to run predictive analytics. 500 600 700 customers – and looking at the right historical data – is enough to yield some results.
All right. Next up is clustering and customization. Whereas predictive analytics was a form of what we call “supervised learning” – where you know what you’re looking for Clustering and customization is actually a form of “unsupervised learning” It’s throwing a lot of data at the problem and asking the machine learning algorithm to find the patterns for you. This is an important part. In marketing and growth we use it to identify patterns to find the characteristic that allow us to segment our different customers.
What are the main characteristics that are important to differentiate my customer base. We call this data-driven segmentation. Whereas we used to sort of guess what the main characteristics were of our different customer segments…now we’ve got an extra vote in the room in the form of machine learning. The recommendation engines are usually built through a mix of the two that we saw just above. A mix of unsupervised learning and a mix of supervised learning.
We call it hybrid models Now with the maturity that we’ve seen above in supervised learning and unsupervised learning, companies have started to use machine learning to build a recommendation engines whereas before we used to use “if-then-that” statements. Although we hear a lot about them in the press like Netflix’s recommendation engine or Amazon’s recommendation engine.
You also see that many ecommerce, content media or transactional companies aren’t actually using them yet and they’re still building recommendation engines by hand. OK now let’s go to number four Natural language processing or NLP, as we call, it is basically asking computers to understand and sometimes reproduce human language. The applications in marketing and growth are also quite interesting. Although they’re not that mature yet we currently use natural language processing for things like sentiment analysis: understanding what customers say about us our product or brand or our competitors.
We can also use it to uncover an indication of how a customer is currently feeling right now on a chat or on an online forum. OK next up is psychographic personas. Demographic s egmentation and behavioral segmentation have a new friend and this friend is called psychographic segmentation. The psychographic is a mix of your personality your interest your attitudes and your behavior. This is a field we’re extremely excited about and we’ve only scratched the surface so far.
The way I like to summarize it is that if you can understand the psychographics of your customers you won’t understand “who they are” but actually “why they buy” which is a great way of delivering the right message and the right product to the right people. And this is different to the clustering I mentioned before
because in this case we’re actually using machine learning to uncover the psychographics of our customers to discover what are the personality traits of our customers are.
Finally the last one at the bottom of our chart is image recognition.
This is actually the one that’s making the most waves and it’s getting the most press on the product side of things Self driving car, facial recognition the stuff you hear about every day in the news For the marketing and growth side we’re still currently exploring applications. Few companies have actually started to
apply this for growth and marketing. We’re super excited about image recognition for marketing and growth but we don’t believe it should be one of your first steps
In what order should you start applying this?
Well ever since we started training machine learning for marketing and growth we’ve actually put these two at the top as the “must-haves” secondly we have the “should haves” once you’ve already covered the “must-haves” and then finally the “nice-to-haves” once your company is more mature with AI and machine learning. Of course depending on which question you want to answer, you’re going to have to start collecting data to be able to answer that question.
Now of course this doesn’t apply to every single business All businesses are different and maybe the order is different for you, but we hope that this can help you in the first steps to applying narrow Artificial Intelligence and Machine Learning to your business and grow.
Web enthusiast. Thinker. Evil coffeeaholic. Food specialist. Reader. Twitter fanatic. Music maven. AI and Machine Learning!