Hack85.Measure Potential Customer Value Using Recency and Latency

Hack 85. Measure Potential Customer Value Using Recency and Latency

The most powerful predictor of future purchases is the measurement of how recently the last purchase was made. Recency and latency are two very powerful metrics for predicting future customer behavior and business success.

As with current value, the repeat visit percentage of a segment is an OK predictor of potential value, but it's kind of a blunt instrument. When you want to start really homing in on defection behavior and potential value, and taking specific action to increase profits, there are special metrics you should use.

Recency is a potential value metric commonly used for predicting customer defection in a business where the customer is in control, which probably includes all the free content and branding models, most commerce models, and some self-service models. A special form of recency called latency is often used in businesses where orders and contacts have a defined "cycle," including those with a defined sales process, subscription-based businesses, and for businesses selling durable goods or high-ticket items. This would include many of the lead-generation and paid subscription content models, as well as some commerce models. We're going to look at both of these retention metrics and how you can use them to rank the likelihood of visitor segments to defect.

6.6.1. Recency

Let's say I offer you a bet. I want you to choose from two customers the one most likely to continue purchasing in the future. The customers are very similar to each other, and each has a lifetime value [Hack #84] of $3,000. There is one difference though: the last purchase of one was over a year ago and the last purchase of the other was in the past 30 days.

Which would you bet is the most likely to buy in the future, and therefore, has the higher potential value?

Most people would select the one that has purchased most recently, hands down. Who knows what happened to the other one? At least the one purchasing more recently has demonstrated they are still in the game.

Customer defection in consumer businesses is usually measured in terms of how long it has been since the customer has had contact with the company, because the longer it has been since you had contact with the customer, the less likely it is that the customer is still a customer. The span of time since last contact is called recency.

Think about your own behaviorhobbies you used to have, restaurants you used to go to, and video games you used to play. How did it happen that you "defected" from those activities? The time between instances of engaging in them grew longer and longer until you stopped completely. That's a defection. And you can predict defection by looking at patterns of recency.

6.6.2. Use Recency to Drive Revenue

A good place to start with recency is to simply determine the "average recency" of your visitor or customer base by following these steps:

1. Determine, on average, how long your visitors wait between making purchases.

2. Once you have an average, create two groups: those with last visit or purchase at the average or more recent than average, and those with last visit or purchase less recent than average.

Customers with last contact dates at the average or more recent than average are the most likely to still be customers. Customers with last contact dates less recent than average are the most likely to be in the process of defecting or are already defected customers.

Don't take my word for this; you can prove it to yourself. You probably have a newsletter or special offer you send to customers and visitors. Flag visitor segments as more recent than average or less recent than average and send out the email. When you look at responses or visits back to the site, you will find that those who are more recent than average have a response rate 3 to 10 times higher than those less recent than average. Why? Because those less recent than average are in the process of defecting; they are your future "former" customers or visitors.

And because they are more likely to defect, they have lower potential value. For example, consider the following table of average visit recency:

Table 6-4.

Visitor or customer source	Average visit recency
Average for all visitors	8 days
From search engines	3 days
From banner ads	14 days

Here we have the average visit recency of the overall web site and two specific sources of traffic: search engines and banner ads. Search visitors are more recent than visitors coming from banner ads, and are more recent than visitors to the site as a whole.

What does the info in this table mean? Let's say search engines and banner ads generate visitors of equal current value; you believe a dollar spent on either ad medium is equally profitable in terms of the value of the visitors generated. But the reality is every dollar you spend on search marketing works much harder than every dollar you spend on banner ads, because search generates visitors with higher than average recency. In other words, the potential value of search visitors is higher than that of banner ads.

You can make these kinds of visitor segment comparisons using a slew of characteristics. The ones listed below are generally the most significant for differentiating potential value segments:

• Media used to acquire the visitor/customer, including the specific search engine and keyword phrases use to find the site

• Offer you made on the initial visit

• Ad copy used to present the offer

• Content areas visited

• Products or categories purchased from

This list might look familiar to you: it's the same list of characteristics we used to look at current value. The same characteristics responsible for creating current value are often strong predictors of potential value. Hopefully, now are you are beginning to understand how powerful these particular characteristics are?

6.6.3. Latency

The basic recency model above works best when there is completely free will on the part of the customer to make decisions. In some businesses, there are external forces or cycles affecting customer behavior. For example, in many business-to-business sales, there are a lot of process-defined sequential steps that have to take place. Many high-ticket sales in general tend to be for items considered "durable goods," which are replaced when they wear out or on some "cycle." In these cases, a related metric, latency, may make more sense to use than recency to determine potential value.

Latency uses the time between customer contacts as a reference point, rather than the time since last contact as recency does. You can calculate latency two ways:

• Look at the average time between visits or purchases instead of the date of last visit or purchase

• Determine the number of days between first visit and first purchase or conversion

Like recency, the longer a visitor or customer does not fulfill the expected behavior, the less likely it is she ever will. If the average visitor segment converts to a lead 10 days after first visit, and you are looking at a visitor segment that has diverged from average, taking an average of 15 days to convert to lead, the segment with 15-day latency has lower potential value than the average visitor segment with 10-day latency.

When you see the behavior of a particular segment diverge from the average behavior of other segments, you get a "tripwire" event. Visitors in the segment are not behaving as expected given the behavior of visitors in the other segments; this likely means a challenge or opportunity with the visitors in the segment. This divergence is like an alarm or flag; it is telling you to pay attention and find out what might be going wrong (or right). Since the calculation of latency is very simple, and the diverging behavior is easy to spot, this type of tripwire is an ideal candidate for "lights-out" or automated rules-based visitor retention/value improvement campaigns.

6.6.4. Use Latency to Drive Revenue

Latency is used much the same way as recency: use the average time between events as a guide, and look for segments with higher-than-average and lower-than-average latency for a particular cycle or step. The average time between actions is the tripwire; any segment taking longer than average to progress to the next step or cycle is likely beginning the defection process. The longer the segment postpones completing the step, the more likely defection becomes.

As with recency, certain sources, offers, copy, content, and products will create visitor segments with average latency either above or below the seven day tripwire, and you should take action to adjust marketing or content appropriately. If your subscription offering is expensive, it might be worth it to be proactive in finding out why somebody who pays a lot of money for the service accesses it less regularly than the average subscriber, and take action to retain the segment. Latency is highly predictive of defection in cases where a regular cycle is expected.

Jim Novo and Eric T. Peterson