We’re focusing on propensity modelling because it’s a tool that digital professionals can really benefit from, and we’ve recently discovered that whilst many are familiar with the term, there seem to be some unknowns around the details. 

Therefore, we thought it was time to lift the lid on propensity modelling so you can make a more knowledgeable decision on bringing this sciencey stuff into your working lives. (Safe to say, we think you should).

 

 

What Is Propensity Modelling?

A quick online search reveals propensity modelling as “statistical approaches to predicting the probability of particular users, say customers or leads, performing certain actions”. That all sounds rather dry. Think of a technique that could predict the probability of yoga practitioners drinking herbal teas or customers using a chatbot on your website instead of calling customer services.

Propensity modelling is a big deal. Predicting the likelihood of someone doing something is a powerful tool for marketers and sales professionals in particular, but also for other business functions too. They can use it to improve campaigns, make more effective decisions, and improve customer satisfaction. 

McKinsey has reported that 71% of consumers expect companies to deliver personalised interactions nowadays, reinforcing our view that propensity modelling is a must-have digital tool for successful companies.

You can argue that propensity modelling is even more important nowadays. Many consumers still struggle with the rising cost of essential items and must cut back on non-essential purchases. In 2023, the global state of consumers was described as unsettled. Our question would be, have things changed in 2024? We don’t think so; hence, propensity modelling seems a great way to combat consumer instability.

One industry that has spotted this is publishing. Publishers went big on propensity modelling after they fell victim to the great unsubscribe when people culled their number of subscriptions after a peak during the pandemic. Publishers like Mediahuis in Belgium have used it to help retain their user base. When customers showed a high propensity to churn, they received marketing phone calls from Mediahuis, which increased retention by 14.17 per cent in just three months.

 

The Typical Use Cases For Propensity Modelling

Using propensity models, companies can address areas such as:

• Enhancing user experience – Through propensity modelling, organisations get a better understanding of customer tastes and behaviour. Using these insights, they can personalise experiences to boost satisfaction and loyalty.
• Improve conversion rates – Models can predict the likelihood of a customer making a purchase based on data like past purchasing behaviour and browsing history. Businesses can use this insight to target the people they see as most likely to convert.
• Reduce churn – Using this model, companies can predict the probability of a customer terminating a relationship. Remember the publishers? Using this data, they knew where to direct the marketing calls.
• Maximise responses – Called propensity-to-respond models, these statistical tools help predict the likelihood of a customer responding to a marketing campaign or promotion. For instance, they can determine the probability of a customer clicking on an email.
• Measure lifetime value – This modelling helps companies predict the amount they will likely receive from a person over their lifetime as a customer. It identifies the most valuable ones, showing where to focus resources.

What are the benefits of propensity modelling to businesses?

In terms of how propensity modelling impacts the bottom line, here is a list of some of the key business benefits:

• Increased ROI – Companies often perform extensive testing to maximise returns when developing new products and services. Propensity modelling analyses data that can refine testing procedures, accelerating progress towards the product and service with the best ROI.
• Reduced costs – Propensity modelling benefits decision-makers, giving them insights to make better choices. For example, data from a propensity model can help focus marketing campaigns, making them more targeted and cost-effective. 
• Saved time and budget – This third benefit is closely related to the preceding two. It gets to how propensity modelling provides data that can help companies save time and money during development. Compared to real-time testing, which is time-consuming and expensive, computer-driven propensity modelling is less costly and faster. 
• Improved customer satisfaction – Customer satisfaction is a primary focus for businesses in these unsettled times. Propensity modelling can help pinpoint customers who aren’t particularly satisfied and could be thinking of leaving. This allows businesses to turn the situation around and improve satisfaction scores.

 

Businesses using propensity modelling

Industries from manufacturing to banking, telecommunications, and utilities have adopted propensity modelling. It has helped them develop websites and services.

For instance, Mitsubishi Motors has successfully used propensity modelling to predict the likelihood or propensity of consumers completing the build-and-price tool for its website. Vodafone has used propensity modelling and other analytical techniques to identify enterprises and customers that would benefit most from 5G. In the case of EDF Energy, propensity modelling was used to reduce churn levels successfully.

 

 

How to start propensity modelling

The first thing to say is making friends with at least one data scientist is a smart move. We’ll lay out some steps to get the ball rolling, but with the warning that you will need someone with a mathematics background to hold your hand. 

So propensity modelling in 5 simple steps would be as follows:

1. Gather your dataset to use for your model
2. Create a propensity model for your chosen use case
3. Explore and validate your dataset
4. Configure and train your model
5. Start predicting using your model

Up to this point, data scientists (like your new friend) have always been the go-to for propensity modelling. But you won’t be surprised to learn that they increasingly share the spotlight with artificial intelligence (AI). Indeed, the demise of data scientists is being openly discussed. 

The reality is that AI has transformed the accuracy of predictions based on existing data sets, accelerating the analysis process and uncovering trends and anomalies faster than traditional models. What’s more, off-the-shelf solutions are available to power your propensity modelling ambitions, simplifying everything tremendously. That said, these solutions may meet the needs of smaller companies; however, a commercial offering may not meet your business requirements for larger businesses. It’s the typical trade-off you’d expect.

 

 

What technical capabilities do you need for propensity modelling?

In our experience, many companies lack the resources to create their own propensity models but have reservations about the ROI of commercial solutions. They feel stuck.

At Ipsos Jarmany, we have propensity modelling capabilities which can help many businesses overcome the hurdles that seem to be holding them back. Our teams of data science experts can create tailored propensity models using machine learning algorithms that will track variables and predict the likelihood of a particular event occurring.

Our flexible approach ensures customers gain the right service to meet their business needs. Plus, our focus on business value helps the companies we work with achieve the ROI they need.

Start the conversation on developing your propensity modelling practice by getting in touch with us today.