Using Predictive Analytics in Customer Segmentation

Using Analytics in Customer Segmentation

In the 90s and early 2000s, when somebody spoke about personalised digital communications, they probably meant they got an email with their name on it or the customer service agent on the phone knew how old they were. There was a point when a barista at Starbucks asking your name was the height of personalised sales.

However, with the exponential growth in technology, artificial intelligence (AI) and connected devices (known as the Internet of Things) as well as the vast amounts of available data in the ecosystem, we are moving into a world of hyper-personalisation, far beyond what digital marketing strategies have traditionally focussed on to segment their customers.

Traditional segmentation models might divide customer into equal groups based on similar profiles such as age or occupation type. Marketing used to assume that all people of a comparable age group had the same needs and wants. However, in life, just because people are in the same age bracket or the same gender, we know that they are far from identical and predictive analytics looks to bridge the gap between data and segmentation.

It is probably a good idea to take a step back and think about how companies can create products that people want to buy. Looking at an every day product like the iPhone, nobody knew they needed a touchscreen device until it was released but there would be limited value in offering that to absolutely everybody because there are some segments that won’t ever want it. One of the best modern-day examples is Netflix who recommend the exact movies that every individual customer will want to watch. This ensures a lower advertising spend through only approaching those with a propensity to purchase, a more fulfilling customer experience as the promotion is relevant and overall, an incredibly successful product.

Predictive analytics using machine learning

Predictive analytics comes from an application of AI known as machine learning. In its simplest form, this is the process of training a machine or computer to complete a task that a human would otherwise have carried out manually. Some have compared machine learning to how a child might learn between right and wrong.

For instance, we would provide experience (historical data) as to what is correct. This might be defining whether a photograph is or is not a picture of a cat. The child or machine would be given countless images and told which were cats. When a new image is presented, using all that historic data, the machine will be able to make connections and work out what the image is without being told.

At the point a machine can make decisions without human intervention and training (known as unsupervised learning), we have truly artificially intelligent applications and predictive analytics can really come into its own.

A common dilemma for digital marketers is working out which customers are likely to respond to an offer or advertisement. Where teams tend to have restricted budgets, it isn’t possible to spend thousands on Pay-Per-Click (PPC) campaigns, Facebook advertising or bulk email so there needs to be some form of predictive analytics in place if businesses are looking to be competitive.

Predictive analytics is all about forecasting the likelihood of an event occurring. Pretty much everyone has been in a situation where an ad has popped up on Facebook based on something they had searched for on Google or visited a website about. This isn’t just a coincidence. Digital marketing teams are using data to reach out to you via the right channel at the right time because their predictive models have forecasted an outcome. Instead of standard customer segmentation where every 25 year old on Facebook gets the same message dropped to them, each 25 year old gets a message an the exact time that the modelling has suggested they need it.

Putting it all together

Putting all of this into common business key performance indicator terms, historic data might tell us which customers provide the best return on investment (ROI). When a new customer comes on board, if their behaviour and demographics match those high performing customers, there could be a case to increase spend on nurturing that customer as opposed to one whose attributes demonstrate a low ROI. Predictive analytics makes marketing planning than ever before and is essential in an ever growing digital universal where billions of devices are now connected. The scalability of Big Data and AI is almost infinite with developments in cloud, edge and quantum computing and the likelihood is that predictive models will only continue to become smarter.