September 07, 2010

Digital Profiles: A New Way To Look At Segmenting Buyers

How many of you find segmenting customers utterly unsatisfying?

You go into your favorite Web Analytics package, and you profile conversion rates across prior visitors. Well, that's fun for about five minutes, then ... zzzzzzzzzzzzzzz, you're sleepin'.

Or maybe you are a cataloger with an RFM scheme from 1987. You know that 7-9 month 1x $100 - $149 AOV customers responded at a 2.94% rate in your August catalog. What does that actually tell you about your customer? Be honest with yourself ... what does it actually tell you about the customers that make up that segment?

What about the OLAP/cube-based wonk who takes twenty-three unique attributes and slices and dices on those attributes ... which is fun, except for the fact that most business leaders can handle one or two or at most three dimensions at one time before becoming completely overwhelmed.

In my time as a consultant, there's two consistent themes that recur:
  • Web Analyst or Business Intelligence Analyst: "We have all of this data and nobody at an Executive level ever listens to what we say. We share our results with them, and they just stare at us, and then they make decisions on gut feel and instinct. Why do they do this when we have all of the data and metrics to prove our point?"
  • Business Leader: "We have all of this data, and nobody can make any sense of it in an actionable manner. It's like these analysts are speaking a foreign language. I don't understand how customers interact across all of my business units, and my web analyst and business intelligence analyst can't explain it to me in an actionable way."
In the Fall of 2009, I decided that I'd heard each end of the argument too many times, that maybe it was time that somebody did something about the problem.

For the past twelve months, I've been working with a concept called "Digital Profiles". I like to think of Digital Profiles as "Modern Segmentation for Savvy Marketers". For the past eight months, I've thrown a "Digital Profile" component into new Multichannel Forensics projects, for free ... sometimes the embryonic product yielded outstanding results ... sometimes, the results didn't really resonate. That's ok when you're providing something for free.

Now, in the Autumn of 2010, I finally have a systematic process that solves a key business problem.

In the next several weeks, I am going to introduce you to the concept of "Digital Profiles", a methodology that segments customers into unique groups that exhibit similar behavior across a wide variety of dimensions.

In our discussions, we'll explore a multi-channel retailer with about 100,000 twelve-month buyers, a retailer that markets via the following channels:
  • Print.
  • Online.
  • E-Mail.
  • Search.
  • Social.
  • Mobile.
  • Retail Stores.
This retailer has nine merchandise divisions (think about tabs running across the top of the website, nine unique areas of merchandise).

This retailer categorizes each zip code as not being in a store trade area, being in one store trade area, being in two store trade areas, or being in three-plus store trade areas.

This retailer categorizes each zip code on the basis of Zip Code Forensics.

This retailer tracks customer recency, frequency, and monetary value.

In other words, this retailer has too much data! This retailer does not have a systematic way to classify customers based on combinations of behavior across physical channels, advertising channels, merchandise divisions, and geography.

I will show you that this methodology combines customer activity across all of these variables, yielding sixteen actionable customer segments. The customer segments can be imported into offline campaign management, execution, and analysis. The customer segments can be imported into your Web Analytics solution, allowing you to analyze customer behavior in new and exciting ways.

In fact, if I were a product manager at Unica, Webtrends, Coremetrics or Omniture, I'd be paying really close attention to this series of posts, because Digital Profiles can be used to unlock complicated customer behavior that normal software applications struggle to illustrate to the end user.

Tomorrow, we begin by illustrating how to generate the data necessary to build Digital Profiles.

September 06, 2010

Page Increases

Recently, one of our loyal readers asked about adding pages to a catalog.
  • "Don't you have to mail a big catalog in order to make decisions about whether you can have a small catalog? In other words, you have to mail big catalogs, they allow you to understand how responsive customers truly are, correct?"
Turns out you can use a small catalog as a base.

Pretend you have a 64 page catalog that generates $2.00 of demand per customer. What would happen if you had a 124 page catalog instead?

Use the square root rule to guesstimate the outcome!

(124 / 64) ^ (0.50) * ($2.00) = $2.78.

Now, simply run your profit and loss statement on each outcome, at each circulation depth, and decide for yourself what the right approach is ... 64 pages, 124 pages, or any other page combination!

September 01, 2010

Classic Multichannel Forensics: The Book Industry

Read this article in the New York Times about reading books the traditional way vs. Kindle vs. the iPad.

This is classic Multichannel Forensics on display, as individuals straddle the fence between the old and the new.

In fact, replace books with "catalog" and replace Kindle/iPad with "online", and this reads like a catalog story from 2002.

August 31, 2010

Analytics Fail

If you follow me on Twitter, you already know about my mistrust of services like Groupon ... no, not a mistrust of Groupon and their business model, but rather, a mistrust of the marketers who see a one-day spike in promotional sales and feel like they've grown their brand in a healthy manner.

The failure, of course, is the analytics used to measure promotions.

You see, more than 95% of what we analyze surrounds a day, a week, a quarter, a season, a year, an item, a department, a division, or a campaign. Marketers, in particular, want to know how a promotion worked.

"Worked" is a wonky word.

Take the Chicago Bagel Authority (thanks @choltan!).

Most analytics tools measure if we sold a lot of merchandise on any given day. It's easy to find a report that says we sold $x on Monday.

It's not easy to find a report that says we generated $y profit on Monday, is it? Seriously, go ask any person in your company if there is a report that tells how much profit was generated on Monday.

Seriously. Go ask somebody.

Last week, I asked 1,970+ followers on Twitter to tell me if they measure the percentage of customers who purchased last year and then purchase again this year. It turns out that this is one of the most important metrics you can measure.

Not one individual said they measure customers in this way.

Every day, we experience an analytics fail. We measure what is easy. It's easy to measure sales, so we spend our time optimizing sales growth. It's hard to measure profit, so we generally ignore it.

In Web Analytics, we've trained a generation of analysts to look only at what Google tells us to look at. Did you ever stop and think about that, the ramifications of a generation of individuals who look at what Google tells them to look at, a generation of individuals who haven't written a line of code, code that allows them to investigate and answer their own questions?

Back to Twitter. When I expressed my frustration with services like Groupon gouging retailers (like Gap) of all of their gross margin, the analytics and marketing community fired back with every conceivable marketing theory.
  • "They will make up the gross margin dollars in repeat business."
  • "Not all Groupons are redeemed."
  • "The promotion increases the number of new customers, and don't you always talk about the importance of new customers?"
  • "You obviously don't think like a retailer, it's about getting butts in the store."
  • "Gap will make money by increasing comp store sales which increases market share which drives up the price of stock which increases shareholder value."
When I asked our community on Twitter to provide the customer metrics that prove their theory, Twitter went silent.

Analytics Fail.

We simply don't measure what matters. We almost never look to see if the customers who purchased via a Groupon promotion ever decide to repurchase. We almost never look to see if we generate profit when we create spiffy marketing promotions ... and if we are one of the rare few who measure short-term profit, we seldom follow-up with measurement of long-term profit.

We act so smart, and so advanced with our trendy, real-time analytics.

And in doing so, we damage the businesses we work for.

August 30, 2010

Lenser Marketing

Some of my competitors tell folks not to work with me, that I am all about cutting circulation.

Now, honestly,
I only want for your business to be more profitable. That's it. Eight times in ten, you'll be more profitable by mailing the majority of your housefile fewer catalogs, reinvesting the cost savings in customer acquisition or online marketing.

But not every time.

One of the firms that competes with me is Lenser Marketing. They recently published a report from Matt Morton that shows, via A/B testing, that contacts at 4 week intervals are better than contacts at 5 week intervals. Click here to read the article.

Any result that is verified via A/B tests and results in your business being more profitable is a good result! You are well served to work with any competitor of mine that validates results via testing.

Summer Segmentation: It's Complicated

By now, you've read a full summer's worth of posts on Segmentation.

Oh, the memories!

On a one-by-one basis, you can segment customers on the basis of the attributes illustrated over the past three months. It's big time fun, and you'll make a bunch of bacon in the process.

But life is more complicated than that, isn't it?

What about the online customer who buys after receiving direct mail that drove the customer to Google to search for the item that the customer purchased, causing you to pay twice for the right to encourage this customer to order merchandise?

What about the online customer who buys from only one merchandise divisions, and the offline customer who buys from that merchandise division and three others? Is there a difference between the customers?

In the real world, there are dozens of dimensions of complexity, dimensions that make it really hard to use a traditional segmentation strategy to understand customer behavior.

This fall, we transition the discussion from Segmentation to Digital Profiles. Digital Profiles are my way of collapsing dozens or even hundreds of dimensions down to a manageable number of segments, segments that contain unique combinations of customer behavior that lead to actionable marketing strategies.

Join me this fall as we explore advanced segmentation strategies ... what I call Digital Profiles!

August 29, 2010

Dear Catalog CEOs: Matchback Magic

Dear Catalog CEOs:

Here's a tidbit to help you determine if your matchback analytics are failing you.

Have your analysts freeze the customer file as of 8/15/2009. Take some random RFM segment from four years ago (46-48 month, 1x, $100 AOV). Measure online demand generated without catalog mailings, and generated via catalog mailings as identified in your matchback process. Calculate the percentage of demand generated online by catalogs (say 50%).

Now, have your analyst pull the 0-3 month, 2x+, $100 AOV (or any other high-value recent segment). Measure online demand generated without catalog mailings, and generated via catalog mailings as identified in your matchback process. Calculate the percentage of demand generated online by catalogs (say 80%).

The difference in these two metrics (50% for non-recent buyers ... 80% for high-value buyers) is highly correlated with how much your matchback analytics are over-stating the importance of the catalog.

Once you know the value of the two metrics, remove the 30% excess (in this example) from your catalog p&l at a segment level, and then calculate subsequent circulation plans based on this analysis.

Then hide your head in your hands and duck, because you won't like it when you see just how much your are over-circulating, benefiting every vendor in the catalog ecosystem, while hurting your shareholders/owner.

An Ecommerce Brand Is Constantly Reinventing Itself

A year or two ago I received an email ... the Leader asked "what happened to all of those COVID customers, it's like they just disa...