6 Tips for Using Propensity Models to Improve Response and Revenue

Predictive analytics is ambrosia for digital marketers who are trying to optimize “right offer, right person, right time” through their campaign management solution. To make it work, you need a combination of defined personas, a content management strategy, creative asset library, and technology (e.g., it’s built into the segmentation tools of your campaign management or marketing automation software).

The end goal is to automate the offer selection and placement based on analysis and predictive models for your particular customer base. However, every marketer can get started by using pattern analysis in your existing response data to identify factors that lead to purchase behavior. Use that data (even manually) to improve your personas, segmentations, and send more relevant offers.

Recently, we worked with a major U.S. retailer to try to figure this out. The retailer includes propensity models in their analysis to identify the best segments by channel – in store, e-commerce, and digital (email, SMS, mobile apps, online advertising). Propensity is loosely based on an RFM (reach, frequency, and monetary) model, and incorporates past purchase data, online behavior, and social media status. Propensity models also help identify the need for a discount to encourage full price shoppers. In the end, propensity models end up being a new layer of analysis on top of the segments to improve the conversations with high value and ready-to-act people.

Another cool case story is found on this YouTube video. (Disclosure: I work for Teradata, the producer of the clip, but was not involved with this effort.)

Meanwhile, here are five things to keep in mind for your own analysis.

1. CRM Is the New Black

If campaign management is driven by precision segmentation rules, then segmentation is driven by pattern analysis of your data. Predictive modeling is great – as it helps to improve content placement inside of various messaging and to time the content when the customer is “in market.” Don’t stop there. Use predictive models and the learnings from your early tests to develop scoring models that improve segments and personas. This should lead you to placing more relevant offers, but also to identifying your best customers, prospects, and upsell opportunities.

2. Collaborate to Delight Customers

Improving segmentation based on predictive models is great, but it gets you no practical results unless the CRM and analytics teams partner with the creative/content teams. Collaboratively, identify the top-level segments differentiated by how customers buy. Then, within each of those segments, target offers based on demographics, job title geography, past purchase, and other insights from the predictive models. Sometimes the optimization is time of day for an email message. Sometimes it’s the offer. Sometimes it’s which offer goes on top or is referenced in the subject line. Subscriber satisfaction will only be complete when all parties collaborate together – and share in the learnings.

3. Think In-Channel Optimization

Analyze marketing effectiveness by channel to reveal trade-offs between media investments, and to improve ROI calculations.

4. Think Cross-Channel Optimization

Customers don’t differentiate your brand by channel, so don’t penalize them for doing so. Break down the silos. Manage offers across channels, optimizing email, SMS/mobile, and social opportunities, as well as offline or in-store recognition.

5. Marketing Data Has Experienced a Makeover

Over the past few years, marketing data has become unstructured and multi-structured, sourced from disparate locations, many of which are not owned by the marketer and vast in volume. Traditional digital marketing data is structured – it easily fits into columns and rows. Social data and online web behavior data is not – it usually consists of long strings of characters without much common structure. Clickstream data is endless; social networks generate tens of thousands of unique data strings each hour and even blogs can generate up to 10TB of data a day – roughly the equivalent of 2 million e-books. Marketers need analytics and segmentation tools that can do more than just capture this data. We need to tap into insights from this enormous bank of often “raw” data. Look for technology vendors and statisticians who can work with graph, time-series, and path analysis – and can work fast on large data loads.

6. More Is Not Necessarily More When It Comes to Insights

The risk of making bad marketing decisions based on weak analysis or poor information is exponentially higher due to the vast volume and diversity of data types available to translate into actionable insights. Make it part of your project goals to identify what data should be retained, what can be thrown away or ignored, and how to avoid duplication.

In another customer story, a B2B technology marketer combines lead scoring and lead nurturing for an experience that moves high-value prospects through a prescribed funnel. The scoring identifies how to dispense premium content and event invitations, and the nurturing enables exploration during the critical research and discovery phases of the sales cycle. Sometimes these cycles can last years – with a big payout. One big learning: allowing prospects to be released to the sales team ahead of their readiness can destroy carefully laid plans. Talk about the need for discipline!

What is your story? Share with us your learnings. Is propensity modeling on your roadmap? What other challenges will you face?

Related reading