Risk Analyzer Primer

Why use the risk analysis

The goal of risk analysis is to identify any potential defectors as early as possible, so you can market to them to keep them active, but it can also help you identify your best customers.

It’s likely that among your customers are some who won’t buy from your company again.  Such customers are defectors – their last purchase was longer ago than your self-defined inactivity period.  Many companies label customers who have not made a transaction in the last twelve months as “inactive” or “defector”.  Each defector represents a lost revenue opportunity, so successful companies try hard to minimize the number of customers who defect.

Loyalty Builders has developed and tested (over several years and many customers) algorithms to identify potential defectors.  Part of the process of analyzing your customers is to assign to each one their own Risk Probability as of the analysis date.  The Risk Score slide bar is a point-and-click tool to use Risk Probability (and optionally other metrics) to select customers who are likely to fade away.  After you set the probability threshold, a single click can give you a list of which customers are most likely to defect.

The accuracy of the Risk Score depends on the default settings.  When you signed up to use Longbow, you gave us your definition of an inactive customer.  We use that definition to calculate how many of your existing customers went inactive each period, for the last several periods.  The number for the latest period is shown in the dashboard section of the Report tab.  We averaged the number for the last few periods, and set the Risk Score slider to yield that many customers.  You expected to lose that many customers.  Now you know which ones are in the at risk group!

Characteristics of potential defectors

By default, only the Risk Probability is used to select potential defectors.  However the Targeting tab has several other sliders, other metrics, to use as additional filters to enlarge or restrict the selected set of customers.  You can also adjust the Risk Score itself.  The default setting for all the sliders will produce approximately the same number of customers as currently defect.  You can adjust the sliders to enlarge or restrict the set, depending on how aggressively you plan your win-back campaigns or how much money you have allocated for the effort.

The additional metrics fall into two categories, descriptive and behavioral.  Descriptive metrics characterize a customer without necessarily judging their likelihood of defection.  Behavioral metrics can definitely impact probability of defection.  The descriptive metrics are

• Revenue percentile  -- How the money a customer has spent in the previous 12 months compares with other customers; a revenue percentile of 71% means that 29% of the customers have spent more money in the preceding 12 months
• Retention -- Length of time between a customer's first purchase and today
• Recency -- Length of time since a customer's last purchase
• LRank -- Relative percentile rank of a customer compared to the rest of the customer population based on the customer's loyalty score, scaled from 0 to 100

The behavioral metrics are

• Risk Probability – The probability, between zero and 100%, that a customer will not make a purchase in the time between the most recent analysis date and the date when their recency equals the length of the inactivity period (in other words, when they become inactive)
• Purchase Delay -- How many times a customer was expected to purchase between their last purchase and the latest analysis date. Used to detect changes in customer purchase frequency
• Acceration percentile – Where a customer’s deceleration ranks among all customers; a deceleration percentile of 97% means that only 3% of all customers are decreasing their purchasing at a higher rate

Estimating your revenue exposure

After you set the sliders for the metrics you elect to use and click the Predict button, Longbow shows the number of customers that meet your criteria and the revenue you may lose in the coming year should those customers actually defect.  The revenue is calculated by summing the trailing 12 months revenue for each selected customer.  In other words, that is what your potential defector purchased from you in the past 12 months.  We suggest trying different positions for the sliders and using the predictions to see the impact on your business of other values for these key metrics.

Identifying possible defectors is the first part of a campaign to keep them as active customers.  The next step is making a relevant offer to them using the Upsell and Cross-Sell services in the Target tab.

Finding your best customers

While the Risk Analyzer is configured by default to find your worst customers, it’s easy to turn it around to find your best customers.  Pick customers whose probability of defection is less than 50%, whose Purchase Delay is less than 2.0, and whose Deceleration percentile is less than 50% (they are actually accelerating).  You could go even further and select customers who have purchased in the last 30 days or whose revenue percentile is in the upper end of the scale.

However you use it, the Risk Score can be a useful tool in selecting sets of customers.