Pumping Up Email IQ With Catalog RFM

In many ways, online behavioral targeting merely formalizes techniques that direct marketers have used for ages. Motivated by the high costs of mailings, catalogers learned long ago how to track customer ordering habits and adjust their mailings accordingly. While email marketers enjoy low distribution costs, a poor understanding of customer interactions can lead to costly opt-outs and missed sales opportunities says Steve Webster, CEO of email services provider iPost. He tells us today how tried and true catalog marketing principles are applied to email in the company's new AutoTarget 2.0 release.

 

Behavioral Insider: How does AutoTarget work to track email behaviors?

Steve Webster: It's fundamentally an analysis engine. It is not meant for acquiring new customers but to improve the ability to get lifetime value from existing customers. It now can be wired directly to your other ESP to drive email outreach. But the behavioral data that is used to perform the targeting is email click and view behavior, which flows into our system. And whenever someone makes a purchase in any channel, that flows back to us as well. Typically, customers set up an automated data feed that FTPs us that data. Every night the system looks at every customer's behavior in those email and purchasing channels for the past twelve trailing months. It compares every customer to every other customer using a hybrid of RFM analysis.

BI: What is RFM?

Webster: It stands for Recency, Frequency, Monetary. RFM analysis was invented decades ago by catalogers. Catalogs cost anywhere from $2 to $10 to drop a book. If you don't get substantial return on actual conversions and purchases you will go out of business. Basically it is human nature that if you recently purchased from a brand, you are far more likely to purchase from them again in the near future. Second-most-indicative of a future purchase is the frequency of purchases you made in the last twelve months. Third-most-indicative is the monetary spend you had with that brand in the last twelve months. Those three variables are indicative of likelihood of engagement with the brand in the near future.

We take the same technique online. Sending email costs almost nothing compared to catalog outreach. But the real cost is when someone doesn't respond to any emails you sent them because they become disengaged because you sent them the wrong thing.

BI: So what actionable data do marketers see each morning?

Webster: The marketing team can see what parts of the list are or aren't engaged and responding. They may stop discounting to the very most responsive parts of the list. In the case of retailers, you stop sending what you have been sending. It's not working. Typically we help them craft a Win-Back offer. It is usually a single very high margin, very deeply discountable blow-out product they can put in front of these folks. If you get them to buy one thing, they are far more likely to buy something else because you re-engaged the customer. Given the low cost of email outreach, you only have to re-engage a fraction of a percent of that unengaged part of your list to generate a dramatic positive ROI, typically 40X.

BI: What is the level of granularity of the segmentation?

Webster: Every customer is slotted in one of 125 different personas, or RFM cells. We take the base and we sort it by recency of interaction with you, if they bought something or interacted with an email message. We sort them from one to five, one being least recent. Within each recency chunk we sort by frequency of interactions, so now you have 25 RF cells. Then within each RF cell, we segment by the monetary spend, one through five. So there are five times five times five cells. They are ordered from 1-1-1, the least recent, least frequent, lowest monetary value customers, to 5-5-5.

Take 5-1-1. It is the very recent but low frequency low monetary value customers. It turns out that all of your brand new email subscribers land in that cell. The marketer sees the actual response of every one of those 125 cells for every piece of email outreach you have done in the last year, who responded strongly or weakly to which kind of offers. You can use this reporting to predict that if you send that same outreach to that same kind of customer that responded strongly, they will respond again.

So, you can look at the response rate of a mailing you sent six months ago, identify the RFM cells that had a strong response, and then pick out of your universe of customers those who are in the same RFM cells. They may be completely different people now, and some may be in different RFM cells now. But the people in the same RFM cells that responded strongly six months ago will respond strongly today. That is why we call it predictive behavioral analytics.

BI: Can you use a client to show how this affected conversions?

Webster: Eastwood Automotive sells automotive restoration products. Their customers have a wide range in levels of engagement and frequency of purchase, because some are dealers and some hobbyists or novices who may buy once.

So how do you effectively market to all these different kinds of customers? They tried marketing based on product preference. But the catalog has thousands of SKUs and coming up with a taxonomy of what kind of purchase is likely to follow off of another is too labor-intensive.

Our tool, instead of being product-focused, is focused on level of engagement with the brand, likelihood of future purchase. They can do a simple segmentation and chop the list in half, part the engaged customers and the other half the unengaged. It is typical that in an email list about 30% you would call engaged and generally respond.

So Eastwood actually sent the engaged part of the list more emails and as a result generated higher demand. They are the people who want to hear from the brand. The unengaged half got fewer emails but they were offers meant to get them to buy. They didn't presume, they just wanted more of what they bought in the past. The overall effect of doing this very simple segmentation was that the demand overall that Eastwood generated went up 18%. Their opt-out rate decreased 14%. And their overall email channel profit increased 25%.

BI: Other than on-site purchase data, what outside data can be integrated in?

Webster: The RFM analysis we are doing relies on interaction with the brand, and different interactions have different scores to indicate level of engagement. Looking at an email message is worth one score. Clicking through is worth a higher score. Making a purchase is worth a much higher score. These things flow into the analysis engine.

But we also have customers who use this for B2B where they aren't actually selling anything directly. Instead, they want to see if customers are displaying 'goal behaviors' on the Web site, where they go to the site and download a white paper. They go to a site and sign up for a user group meeting that is worth another score. The scores flow into the system as well. We call this RFV, where V is value. Different industries have different goal behaviors and ascribe different values to the behaviors their customers have.

It's the very same analysis that catalogers started using decades ago. It can be used online to let marketers target exactly who in the customer base is engaged in responding and then who is disengaged and the target them appropriately. The overall result is far more efficient and with higher ROI outreach in the email channel.

Recommend (6) Print RSS