The Istobe eCommerce Analytics Blog

I recently had the opportunity to sit down with one of the marketing professors at MIT’s Sloan School of Management to discuss the current state of marketing analytics. One of the many topics that came up during our discussion was why so many companies are still basing their marketing strategy on RFM. For those of you unfamiliar with the term, RFM stands for recency, frequency and monetary value. It has been used in direct marketing for the last 30 years and continues to be the basis of many “rule” based email and direct mail campaigns. Why? Well, for starters, it’s easy for many marketers to understand: customers who bought recently and have a history of buying large amounts often are more likely to purchase again in the future. Unfortunately, RFM fails to consider many of the factors necessary to truly evaluate the profitable of any one customer - mainly, the cost of acquisition and the cost of customer service and retention.

Does it matter? Absolutely. Way back in 2003, Rajkumar Venkatesan and V. Kumar published a paper highlighting the benefits of using CLV over RFM. They showed that the net profits from the top 5% of CLV-ranked customers were 1.6 times the net profits of the top 5% of RFMers. Additionally, they found that using CLV to better target customers increased profits by almost 67%.

So with increases like that, why are so many companies still using RFM? Well, to be honest, most companies still don’t know any better. However, as the global community gets more competitive, savvy marketers are beginning to look past RFM to make better use of their data through more advanced data analytics. And the good news is that the field of marketing analytics continues to provide us with better ways to analyze data every year. While CLV continues to be one of the best ways to target customers, research by professors like Pete Fader at Wharton and Dipak Jain at Kellogg have given us models that more accurately forecast number of purchases and retention rates of customers for non-contractual businesses. Recent papers have also focused on enhanced forecasting of the migration of customers from direct mail marketing to email blasts (check out Customer Channel Migration by Ansari, Mela, and Neslin).

With the rising cost of direct mail and the waning interest of customers through email blasting, it is clearly the time to improve the effectiveness of marketing to improve the overall profitability of the business. While I’ll admit that RFM was a great marketing tool in the past, there have been so many advances in marketing analytics since then that it’s time to move on.

The knock on B2B data mining has always been that there isn’t B2C-like data available. Instead of multiple transactions that give us customer behavior patterns, we have company demographic information (industry, company size, revenue), some information about the person from the company who we’ll deal with (position/title), and where that person came from (lead source). It’s not behavioral data, which we know to be inherently better as a predictor than demographic data. But some data is better than none, right?

And we can certainly create transactional data that gives us some behavior pattern. If we throw in the contact schedule - the touches - from your company’s representatives, don’t you have a transactional pattern of both buying and non-buying customers? Coupled with the demographic data, you can drum up a model that predicts how many touches a lead might need to become a client and maybe a best guess at the path that should be pursued with a new lead.

More to the point of this post, this is the great thing about click-through data: it has a transactional quality. In fact, it just might be the transactional data for B2B companies. (Aside: This is also one of the reasons why companies like Omniture are becoming so notable: they provide some behavioral patterns, however small.) If we can combine click-through patterns from the person representing the prospect company with the company’s demographic information, then we might have a real interesting model that determines just how serious a lead is about buying from you and their company’s relative experience level with your product area.

Let me close out this post by refuting two of the main complaints about B2B data and its unsuitability for data mining-based models.

There’s Not Enough Data

Everybody loves data mining when it comes to consumer-focused companies. The vast amounts of transactional data are transfixing. The thinking goes something like this: “I’ve got hundreds of thousands of transactions here so whatever our predictive model spits out must be right.” Well, this may be true. And it mayn’t. But that doesn’t make a model built with less data any less compelling. It just means that one model has more data points. Don’t feel inadequate for the difference. Just make sure that you have data that’s important to the business problem you’re trying to solve. For example, if you want to know the next-best product for newly-minted customers, then you’d better have a solid set of second-time customers who bought a bunch of different products. Do you need thousands of these second-time customers? C’mon.

Missing and Bad Data

Isn’t this a reality everywhere? Even consumer-focused companies (with hundreds of thousands of transactions) have this issue. Oh, and I have a suggestion on what to do with that missing and bad data. Throw it out. Chances are, it will have absolutely no effect on the predictive models, unless of course all of the missing or bad data has a common characteristic that isn’t found in the rest of the data. For example, let’s say you’re building a model that predicts the next software product that a first-time customer might want from your company. Well, if everybody that bought a specific product as their first purchase is missing a zip code, then you can’t very well throw all of those records out. It would skew the model irreparably. But as long as the missing data is evenly distributed throughout the records, don’t be afraid to trash ‘em.

Multichannel retailers are reluctant to stop sending catalogs to customers who primarily order online. There is a good reason for this. The catalog is undoubtedly the impetus that drives many buyers to order online, even if those customers don’t enter a catalog code. But even so, catalogs are not as critical as matchback analysis suggests. Why?

Because no matter the date, most high value customers just received a catalog.

High value buyers are frequent buyers and, as a result, they spend most of their time in the 0-12 month customer file. Customers in the 0-12 month file generally receive around one catalog per month.

At the same time, matchback analysis attributes all online sales to the catalog if the customer received a catalog in the preceding two or three weeks. That leaves little time each month in which an online sale could not possibly be attributed to a catalog. Yet there must be cases in which a received a catalog within the past three weeks but the catalog did not spur the order. Matchback analysis has no way of identifying these cases, which I suspect are pretty common.

Clearly this methodology is faulty, so why does it continue to be used? Call me cynical, but I suspect it has a lot to do with the fact that list vendors have a vested interest in promoting catalogs as a marketing vehicle. Also, it provides some comfort to catalogers with large house files who want to believe their big circulation numbers give them a strategic advantage over internet only retailers. In this way, matchback becomes a fantasy in which the sensibilities of the traditional direct marketer are reaffirmed.

To truly understand how many online purchases are being driven by catalogs, we can explore a different technique: holdout testing. I’ll explain more in my next post.

I just ran across this quote from Will Schnabel and it gave me pause. In Lead Score and Activity Alerts: One in the same?, Schnabel says, “In order to identify the few qualified prospects from the remainder of the inquiries, you first need to determine your ideal customer profile or in other words, what segment is your best target. From there, BANT questions (budget, authority, need, and time frame) help determine the qualification status of the leads.”

So I agree with the last part. BANT qualification is important, especially in the B2B and B2P (Business to Prosumer) markets. After all, no one will ever buy your product without budget or the authority to make decisions. It’s the first part that’s becoming increasingly interesting and more complicated for me. Is the customer profile so rigidly first in this timeline? Or is privileging this activity just another reason that the Marketing/Sales political split persists? Certainly, this is the right way to pursue new business strategically: decide the size and relative need of a target market, build or alter product to address needs in that market, create messaging that speaks to the market about its needs, and then attack the target market.

Well, what happens when that target segment trickles in while a different segment - one not entirely counted on - starts to beat down your doors? OK. If the doors are really being kicked in, your company is likely to notice. But let’s say that this secondary segment buys on par with the target. Sales is likely to notice it first, right? And this is usually what causes the rift between marketing and sales: The marketing department often continues to chase the strategically-desired and well-messaged segments instead of the empirically-best segments.

This is where a data mining approach refines the way we look at segments. There may be unforeseen criteria that ultimately make its way into the segmentation definition. Clickthrough data is the most interesting and most-discussed piece of data in this vein. Schnabel would have you believe that this data determines the relative need of the segment in the BANT qualification. I say, why can’t it be part of the segment itself? A shrewd marketing department might change the amount of information available on the website so that the “heavy researcher” psychographic segment is more fully qualified. Particularly if the heavy researcher is also a heavy buyer.

Many companies are starting to break down the barriers of B2B and B2C. The B2P audience is a great example of this. This means that traditional customer segmentation is becoming more difficult. Customers may be segmented along demographic, business history, behavioral, and psychographic variables all at once. Ultimately, a data mining-aided approach to customer segmentation lets you adapt to customers who find you by breaking with these traditional barriers to make segmentation a fluid discipline.

Ask any direct mail marketer what the most disruptive force in marketing has been in the past ten years and the answer will undoubtedly be the rise of email. Email has many advantages over direct mail: it’s cheap, easy to send, and allows marketers to easily track results to learn what works and what doesn’t.  Despite that, email marketers would do well not to be seduced by the lure of “free” email.

Unlike direct mail where the postage costs are front and center, blasting 100,000 customers with the same message doesn’t cost much more than blasting 10,000. As a result, it can be tough to resist the temptation to add just one more customer segment to the blast list.  Here’s where it’s important to recognize the real hidden cost of email: the opportunity cost of a customer becoming blind to (or opting out of) your email because the messaging is too frequent or not relevant.

Put another way, consider that the marketers job is to nuture the full potential lifetime value out of each customer. A thoughful, well targeted, and patient email campaign can nudge a customer toward realizing that potential. Get too greedy, however, and some customers will tune out forever. If one in 10000 customers tune out, and the average unrealized lifetime value of those customers is $500, then each email actually costs $0.05 plus about $0.01 for actually sending the email. That’s still cheaper than a catalog, but it’s far from free.

To keep the real cost of email marketing down, ensure that you are sending targeted and relevant messages to each customer segment. You can do this by analyzing the buying affinities of each segment and making sure the offers you send are for products that will interest the reader. If you don’t have the time or resources to crunch the data, don’t be afraid to rely on your intuition. The important thing is to remain disciplined about shielding customers from irrelevant communications.

Most of the talk today in B2B marketing is about gathering leads. With Google AdWords going crazy on us and SEO the hot new term, it seems that everybody is intent on generating a boatload of new prospects. Nothing wrong with that. But it does create a problem of scale. Does your organization have the salesforce to handle the explosion of leads? Probably not. Why is that a problem? Because without a lead scoring system, you’re trading quantity for quality by putting some of your best prospects on the backburner. With the glut of new leads out there, it’s easy to see that lead effectiveness is more important than ever.

Sure, most companies have some form of lead scoring in place. But the two most common methods, telemarketing pre-qualification and manual qualification, are expensive in the first case and inaccurate in the second. Manual qualification has the added problem of exposing the divide between sales and marketing. (They often have diverging views on what a good prospect really is.) And scoring is also only as good as the rules that are created. If the scoring system isn’t constantly revamped based on the actual sales wins, then what good is it?

Data mining techniques are different than common scoring techniques because they constitute an impartial observer. Data mining keeps departmental biases and political loyalties out of the equation. More importantly, data mining techniques effectively lop off the worst leads - the leads that clog the sales funnel - simply by projecting a prospect’s likely behavior based on your past prospects’ behaviors. Best of all, data mining techniques can account for hidden attributes that a human can’t. Did you notice that your best prospects visit your web pages in a specific pattern? Well, a well-trained predictive model does. And it’ll mark the clickthrough pattern as more important in determining the prospect conversion probability than how many white papers they’ve downloaded.

Now that leads are on the rise, B2B marketers should focus on increasing the sales win rate by intelligently funnelling the best leads. And this means dropping some of the prospects that likely won’t produce. To weed out my worst leads, I’m betting on data mining over rule-based scoring systems any day of the week.

Do you sell recreational, hobbyist, or athletic gear? Are you looking to acquire high quality customers? Well, as much as we don’t like to admit it, not everything has to be analytics and data mining. Sometimes intuition is enough to achieve our goals.

Consider that most recreational activities have high and low seasons. For example, recreational boaters and pilots are most active during the summer months. Skiiers and hockey players are most likely to be thinking about picking up new gear during the winter (or shortly before).

Our intuition tells us that if someone, say, buys a new pair of skis in the summer, they are either a) well off enough to take a summer ski trip to South America, or b) are incredibly devoted to the sport. Either way, he or she is a likely highly desirable customer.

As a result, we should be willing to spend more to acquire customers in the low season. The low season is the time to crank up the minimum cost-per-click bids on Google or spend a little extra to rent a list. Let your competitors slug it out in the peak activity months, fighting tooth and nail for the dabblers and tightwads. You’ll be sitting back collecting money from your stable of devoted high spending enthusiasts.

In RRW Consulting’s Direct Marketing Blog, Nancy Arter mentions that, in order to properly embrace Green Marketing, direct marketers need to “prove it” to their consumers via their business practices. Otherwise, she says, a green message rings hollow. As an example, Arter states that catalogs and coupons need to be printed on recycled paper as proof to the consumer that your company practices what it preaches.

Well, I’ll take it three steps further. I would argue that Green Marketing means sending as few catalogs as possible to achieve the same revenue targets, regardless of what type of paper those catalogs are printed on. So am I advocating that catalogers move all of their operations online? Hardly. That would be fiscally irresponsible. Catalog sales, after all, are still going strong. And there are demographics that ecommerce will never reach; demographics that include catalog devtoees who will not soon, if ever, buy from the web.

But the shrewd direct marketer should see the greening of marketing as an opportunity to prune their mailing list by using customer data best practices. Current data mining techniques make it possible to predict which customers will never buy from you again, regardless of what types of advertisements and enticements you send them. So why send those customers catalogs at all? We recently ran a study and found that we could save a multi-channel retailer a projected $47,000 a year by using customer models to identify the customers unlikely to buy again. Our recommendation: save the recycled paper by pruning those customers from the mailing list.

Direct marketers who parse their catalog lists using the powerful data mining capabilities available can make a solid case that they are listening to consumer opinion while trimming their own fat. Now, more than ever, direct marketers have the tools at their disposal to send less catalogs while still achieving the same marketing lift. And they can tout their sustainable practices back to their customers. Now that’s a marketing organization that will see green.

Last week in What is Customer Analysis? we found that the first step in a customer analysis is determining customer lifetime value across segments. Armed with this information, we can determine which customers are worth focusing our marketing efforts on and which customers should be “fired.”

The concept behind modeling customer lifetime value is relatively straightforward. We can group customers into segments which behave similarly and then based on historical data, determine how much a customer in each segment produces in profit over the course of his/her lifetime.

One thing to understand with calculating customer lifetime value is that there are many different ways to do it. Practically speaking, as long as you remain consistent in your usage across all your customer segments and across time, you should be ok using any of them.

With this in mind let’s look at one of the simpler customer lifetime value formulas:

Where,

m is the average gross margin
i is the discount rate
r is the customer retention rate

In using this simplified formula we to pick one average profit margin value and one average customer retention rate.

We can calculate m for each segment as

m = revenue - product or service costs - cost of servicing (includes acquisition and promotion costs)

over the course of a period (usually a year).

To find r, we calculate from historical data what percentage of customers in a segment repurchase in the next period (again, usually the next year). We then assume that this will also be the retention rate for subsequent years for the segment. This is generally not the case but that’s a price we’re willing to pay to keep things simple.

Finally, i is the cost of capital (sometimes called a hurdle rate). If you don’t know what your company’s discount rate is, your CFO will likely be able to give you the number. If not, you’ll usually be ok using a value between 8% and 15%.

Using the customer lifetime value formula to rank each of your customer segments will give you a solid understanding of which to court and which to forget about. In particular, any customer segment with a customer lifetime value less than zero is costing your company money. Shed yourself of these customers as quickly as possible!

Anxious to get started calculating customer lifetime value? Our customer lifetime value calculator will help you get started. If you require more precise CLV calculations or CLV calculations on an individual customer by customer basis, consider a solution such as our own customer analysis software.

I feel like June is always a great time to take a look back at the forecasts and predictions made at the end of the previous year to figure out what’s living up to the hype and what has yet to catch on. Many internet retail trends have yet to meet expectations - virtual world advertising (i.e. Second Life) and YouTube marketing are still in their infancy while other marketing areas like personalization and one-on-one marketing are just starting to gain mainstream buy-in. One trend that we believed in at the end of last year that is now being implementing at major retailers across the country is the integration of online and offline marketing campaigns to maximize the effectiveness of cross-channel sales.

Of course, the question is always how best to judge the effectiveness of a campaign that generates sales in multiple channels. Unfortunately, this is not just difficult, it’s becoming harder as many customers are increasingly using one channel to research information about a product while purchasing the product in a different channel. As an example, the number of customers that receive a product catalog and then choose to purchase via the phone or mail has dropped significantly over the past five years. While it may seem like this would make the case for discontinuing expensive catalog mailers, the truth is that catalogs still drive a substantial amount of purchases - customers are using catalogs to inform their decisions about web purchases.
Back in February, eMarketer came out with a great report that showed the inverse is also true. In his “Multi-Channel Retailing” article, Jeffrey Grau writes about how buyers are increasingly using the web to research a product that they intend to purchase in a retail store. In fact, the article estimates that for every $1 that is generated from online sales, nearly $4 is generated from in-store purchases that are driven by online research. Additionally, over 90% of consumers that purchase online with some frequency have used the internet to inform themselves about items they later bought in-store.
So why is this important? Well, first, if you’re selling across more than one channel, it’s important to recognize that even simple marketing decisions may have a greater effect on your customers than you anticipate. One less catalog a year may seem like a good idea to someone with floundering mail order sales, but unless you can determine which of your customers is using the catalog to purchase online, you may be effecting more than just your mail order business. Second, no matter how your organization is structured - whether your online sales site is run as a separate division or if all marketing is a centralized in one department - the only way to truly judge the effectiveness of your marketing campaigns is to carefully track and, more importantly, analyze the data from all marketing related activities in one place (that means using everything from Google analytics to catalog match-backs). Last, while many internet retail trends that were forecast to happen in 2008 may have several years before becoming mainstream, the recent rise in postage and supply costs means we’ll be seeing more and more retailers looking to get the most out of their multi-channel marketing dollar in the coming year.