Wearable fitness technology is a powerful example of using data visualizations as a storytelling device. Avid cyclists, runners and swimmers recite their latest pace splits, heartrate zone thresholds, cadence and real-time VO2 max rate as easily as recalling what they had for breakfast. Apps provide updated pie charts, bar graphs and Venn diagrams with every workout, showing real-time progress (or regression) and providing instant visual analysis. By drilling down into the data, fitness buffs have a better understanding of optimal recovery time, how hard to push or back off on ensuing workouts and where to make improvements. Visualizations inspire a data-driven approach to fitness; seeing is believing – and motivating.
The concept applies to data-driven marketers presented with visualizations of segments, customer attributes, propensity scores or rules that divide an audience for marketing purposes. More than seeing a representation of the data, it’s the ability to interrogate the data – to see how and why a decision was made – that bestows trust in the underlying data foundation, sparking confidence and inspiring marketers to be creative with message and offer design.
A Visual Understanding
Ideally, an audience segment and the resulting visualization should be as distinct and cohesive as possible. A tightly fit grouping or cluster of customers with an easily understood, common set of characteristics enables marketers and business users to derive meaning from the segment. A statistical or visual representation of two or more variables in relation to one another that is both distinct and cohesive (e.g. product affinity vs. demographics) yields the outcome that marketers want, i.e. a framework for establishing a meaningful correlation between data points.
Every marketer planning a campaign wants a firm understanding of how one group of customers differs from another. The deeper the insight – that is, understanding what motivates them, the essence of their relationship with the brand – the better a marketer understands how one group is likely to respond vs. a different group, or how each group best aligns with campaign goals. In that regard, visualizations of clusters, propensity, affinity and other statistics, rules or models through scattergrams, pie charts, bar graphs, colored heat maps or even 3D visualizations become important tools for preparing and executing campaigns.
A Venn diagram, as one example, is typically used inside of defining a rule set for an audience or cluster. Perhaps there are three or more variables being used that contribute to a rule, and the objective is to view an overlap between those variables. A Venn diagram will represent the overlap visually, showing for example how many customers are both high propensity to buy and high lifetime value. The overlaps present opportunities for experimentation by providing marketers the ability to drill down into what characterizes one customer from another, using that knowledge to guide different customer journeys or perhaps even plan an entirely new campaign.
Redpoint rgOne supports all of these ways to visualize and interrogate the underlying data to discover meaningful correlations (see Figure 1). The platform’s Clustered Audiences functionality illustrates how users are able to extract meaning from visual representations. A marketer may present a machine learning-driven model with customer data and ask the model to return a specified number of segments – clusters – while also presenting a drilldown option with a decision tree that shows the reason, at each inflection point, for why the decision was made.
In this case, a marketer or business user might provide a range of variables and a desired outcome, such as an understanding of lifetime value or propensity to purchase, and the model itself finds similarities and common attributes, discovering correlations that might not occur to a marketer or business user of the data.
A marketer might, for instance, be planning separate campaigns for low, medium and high lifetime value customers, with the resulting clusters providing distinct, cohesive groupings and an explanation for how the customers were divided. The clusters might also reveal interesting edge cases or anomalies with the potential for niche marketing use cases to attract high value customers.
Campaign Monitoring through Visualizations
Using data visualizations at the outset to help plan campaigns is a familiar use case. Another data visualization use case is pertinent during an ongoing campaign, centering on monitoring the progress and effectiveness of a campaign.
A marketer may wish, for example, to measure uptick and see the division for how many customers are clicking on Offer A vs. Offer B. What do those customers who click on Offer A have in common? If, at the outset, a supposition was made that a predictive model fed a list of rules and attributes was going to reveal something meaningful about customers, is that supposition accurate? Visual representations such as a scatter graph can measure any number of performance-related metrics – product sell-through vs. offers made, sell-through vs. images clicked, etc. Through data visualizations, marketers can easily analyze results and change course if needed to optimize a campaign.
Similarly, visualizations may also be used for reporting and attribution at the end of a campaign. One example would be a visualization of money spent on ads, emails, SMS messages, etc. vs. the results generated.
As a final note, it should be clear that data quality is of the utmost importance for marketers to be able to trust downstream findings. Data quality itself, then, might be overlaid on a decision tree using a heat map or another visualization tool to help determine if the data itself is fit for a campaign.
This might be an important step if there is a concern that there are differences in data quality from multiple sources. A point-of-sale system might, for instance, be known to be more accurate than data coming in from a distributor. If each of those sources contributes to a cluster, the quality of data might be an important variable to consider if the resulting clusters determine which group of customers receives which offer.
Ideally, every data visualization should provide actionable insight by highlighting clearly defined differences between customers and the reasons for those divisions. The key to a visualization being actionable is that it is built on a strong data foundation; knowing what attributes make up a segment gives marketers confidence that they’re on the right path with regards to the beginning, middle and end of a campaign, and for ultimately delivering each customer with a personalized customer experience.