We often hear about fantastic AI cases and how AI saves the world… But the truth is that AI is all about data! Without data, AI is just another computer code. So how do you get your data in order? How to avoid the pitfalls of costly master data management projects that end up in failure?

We will start by finding value through a use case.  Let us consider an example: a Customer 360 View (as this is the most popular use case everywhere). A case like this begins with analysing the current situation of the customer data. A company that has multiple offerings to customers might have separate CRMs and separate customer IDs in every business unit. In most cases, also the development of IT systems usually starts from a company and product perspective, not from a customer perspective.

If you truly want to understand your customer, you really should be able to see all the offerings and services the customer is getting from you. This is the first step in the right direction. You either have to build a data layer or harmonize your back-end systems to view the customer as one. If you have the courage in your organisation to boost this kind of approach, you will have to face all sorts of difficult political questions, such as “Who owns the customer?” Organisational politics have a big role in change leadership, and the only possibility in a situation like this is to bring data widely available (what we call data democratisation).

Predictions based on customer features or behaviour?

So let us assume the first step is done and your transactional data about your customer has been combined. Your next step is to understand your customer better: for example his or her age, location, living situation, family, and interests. Here we take one step towards analytics or even AI. We might try to cluster our customers based on their features. This is one way to try to understand our customers better.

However, a more common way is to follow the customer’s activities online, and based on this behavior, to find the nearest neighbour. People who have acted accordingly can give our AI tool an indication of the most probable behaviour of this type of customer. This is one of the approaches used in all recommendation tools, such as those employed by Spotify or Netflix. Just remember that this approach works only if you have a big amount of data.

3 tips to get started with Customer 360

Now, let us get back to our original use case: a Customer 360 View and how to get started.

  1. Define your actual need. Why do you want to get this kind of view of the customer? Do you want to predict the churn, find new customers, or increase revenue per customer?
  2. Based on the actual case and viewpoint, select your tools and platforms. Evaluate where your data lays at the moment: is it in a usable format and available for analytics?
  3. When you start getting the results of this case, the ultimate step is to train your teams and start the cultural change towards a data-driven practice.

Even with perfect analytics and information sharing, you have to remember that a culture change is needed to actually change people’s behaviour. To change the culture, we all need to be aware of some basic facts about AI and data: for example the fact that AI solutions are never 100% correct, but they can give you the most likely result within a certain accuracy. Also, the quality of data makes a huge difference on the prediction and the results: the realistically blunt slogan is “Garbage in, garbage out.”

When we have educated our employees and decision-makers to take full advantage of data, we have reached the moment when the magic happens and we can actually start creating value to both our business and our customers.

 

About the Author

Kristiina Söderholm is the Data Leadership Practice lead at Sofigate, experience change leader and PhD of nuclear science.

Kristiina is interested technology opportunities and data. She likes to make data available and demystify technology and data science. Her motto is ”Free fall is an opportunity, not a threat!”

Read these next