Online / Retail Dynamics: Creating Nine Segments

Ok, you've made it this far, congrats!

Now, if you don't like geeky math, here's the place where you move on - there's nothing for you to see here. I'll catch up with you another day.

Once I have my variables defined for the past year (or weighted based on the past "x" years), I plug the variables into a Factor Analysis. Here's the SPSS code I used for a recent factor analysis.

FACTOR
  /VARIABLES
  d r visits03 visits10 visits30 visits99
  m00 m01 m02 m03 m04 m05 m06 m07 m08 m09
  m10 m11 m12 m13 m14 m15 m16 m17 m18 m19
  m20 m21 m22 m23 m24 m25
  price
  cat aff ema nat pai ref soc
  /MISSING LISTWISE
  /ANALYSIS
  d r visits03 visits10 visits30 visits99
  m00 m01 m02 m03 m04 m05 m06 m07 m08 m09
  m10 m11 m12 m13 m14 m15 m16 m17 m18 m19
  m20 m21 m22 m23 m24 m25
  price
  cat aff ema nat pai ref soc 
  /PRINT INITIAL ROTATION FSCORE
  /CRITERIA FACTORS(02) ITERATE(200)
  /EXTRACTION PC
  /CRITERIA ITERATE(200)
  /ROTATION VARIMAX
  /SAVE REG(ALL)
  /METHOD=CORRELATION.

execute.

There are several variables ... "d" and "r" represent percentage of demand in the direct and retail channels ... visits are categorized (visits03 to visits99), each merchandise category is given a weight that sums to one (listed via the m00 to m25 variables). Average price point is included, and I have weighted variables for catalogs, affiliates, email purchases, natural search, paid search, referrals, and social media.

This analysis yields two factors ... essentially, I am reducing the variability in the dataset down from 30-40 variables to just 2 dimensions.

Next, I create a segmentation variable ... I categorize the 2 dimensions based on the values of the 2 dimensions.

compute seg = 0.
if (fac1_1 lt -0.40) seg = 10.
if (fac1_1 ge -0.40) seg = 20.
if (fac1_1 ge  0.40) seg = 30.
if (fac2_1 lt -0.40) seg = seg + 1.
if (fac2_1 ge -0.40) seg = seg + 2.

if (fac2_1 ge  0.40) seg = seg + 1.

This little bit of code gives me nine segments ... the customer can be low/medium/high for the first factor, and low/medium/high for the second factor.

At this point, each customer belongs to one of nine segments. Now is the time where you drill down into each segment, analyzing the characteristics that yield each segment.

We've got it! We're there! We now have a map that tells us how customer behavior changes across segments. Each segment should be targeted differently. Go make something happen!

This is the process I go through. The variables, of course, are different in each analysis, and quite honestly, there is an art involved in creating actionable variables that yield actionable segments.

Ready for your own, customized Online / Retail Dynamics analysis? Good! Contact me (kevinh@minethatdata.com) and let's get started!!

Online / Retail Dynamics: Weighted Attributes

There are many variables that I like to analyze, on an annual basis, in an Online / Retail Dynamics project.

Merchandise Categories: I sum annual demand by merchandise category, then divide the total by how much the customer spent in the past twelve months. This gives me a fraction (0.00 to 1.00) of amount spent in each category. Some clients want two years or five years or all history included. I find this is not an optimal way to analyze what customers purchase - who cares that you purchased a love seat in 2004? In these cases, I weight historical spend ... maybe 100% for 12-month purchases, 50% for 13-24 month, 30% for 25-36 month, 20% for 37-48 month, 14% for 49-60 month, and 10% for 61+ month purchases. This greatly minimizes the influence of old purchases, especially older high-dollar purchases.

Website Visits: Here's a little secret not many folks want you to know - in most of my projects, I'm asked to analyze twelve-month website visitation behavior. Having said that, on average, only website visits in the past 15 - 30 days have any influence on future behavior. Often, I'll create three variables ... website / mobile app interactions in past 30 days, then from 31-90 days ago, and finally, 91-365 days ago. But again, only the most recent website / mobile app interaction matters. Recency is critically important online, folks. Heck, sometimes I'm asked to group website visits into buckets ... 1 visit last year, 2 visits last year, 3-5 visits last year, 6-10 visits last year, 11-50 visits last year, 51-100 visits last year, 100+, that kind of thing. Whatever works for you is fine, just make sure you have a defensible point of view.

Purchases: I like to sum twelve-month (or historical weighted) dollars by channel ... retail, smartphone, tablet, desktop/laptop, call center, that kind of thing. Then I'll divide the totals within channel by total annual (or historical weighted) dollars.

Website Characteristics: Here, I like to categorize activity on a weighted basis ... 100% for 0-30 day activity, 20% for 31-90 day activity, 5% for 91-365 day activity. I'll create 1/0 indicators for all key characteristics (cart, email click-through, referral from Bing, that kind of thing), then I will weight each characteristic by time (100%, 20%, 5% as mentioned above), and create a percentage. The weighting becomes important ... if a customer visited via Bing 100 days ago and Google yesterday, the Google visit is weighted at 100%, the Bing visit at 20%, meaning that the customer has a Google preference at a rate of 100/120 = 83%, while the customer prefers Bing at 20/120 = 17%. On an annual basis, the weightings really help us understand how the customer behaves.

Store Distance: I'll plug 1/0 indicators into my analysis for 0-5 mile bands, 6-10 mile, 11-25 mile, 26-50 mile, and 51+ mile bands. You will learn that visitation behavior changes as customers get further and further away from a store.

Zip Codes: I categorize zip codes by Catalog-Centric, Online-Centric, and Retail-Centric. Behavior in each classification is simply different, and quite interesting! You probably have your own algorithm for categorizing each zip code, so use that.

Tomorrow, I'll show you how I cook this information up - the discussion might get a bit geeky, but that's the nature of the work I'm doing when analyzing Online / Retail Dynamics.

Online / Retail Dynamics: Data Structure

There are two ways that I typically analyze Online / Retail Dynamic data.

If the client is looking for purchase-specific information, then I create a table similar to the one outlined here. Each order is "tagged", if you will, by the channels that were attached to the order. Look at the first purchase. This customer bought in a store. Each column that follows outlines a yes/no indicator for what that customer did in the 30 days leading up to that purchase. In this case, the customer clicked through an email campaign, visited the site via natural search, visited the site via social media, visited the core website, and engaged with the retail app. You can imagine the columns that would be valuable to your business - just tag each order with the channels that impacted that order in the thirty days prior to a purchase, and start analyzing!

I tend to use a different approach in most projects. Specifically, I analyze customer behavior across a full year. The information, across a year, is so much richer and more revealing than the information tied to a specific visit and/or purchase. We learn that customers have interesting, consistent, reliable behavior that is masked by noise in individual visits ... but becomes easy to see on an annual basis.

Tomorrow, I'll talk about the variables I use, on an annual basis.

Online / Retail Dynamics: Competition

When I worked at Nordstrom, we used to read all sorts of interesting content about Neiman Marcus, one of our primary competitors. Most interesting, of course, was the difference in tone.

• Typical Neiman Marcus Comment: "We are out to crush the competition."
• Typical Nordstrom Comment: "We welcome Neiman Marcus into our markets, because when they open stores, we find that our comp store sales tend to increase."

Catalogers know this as well - they've been sharing their most coveted asset - customers - with each other for thirty or more years.

Yes, I get it, your competition is "just one click away". Fine. At Nordstrom, we analyzed all online and retail behavior within the context of the amount of competition a customer faced within a zip code. Online, we couldn't control for this dynamic (though referring URLs taught us that our customers shopped everybody - and credit card data showed us that customers shopped Wal-Mart as much as they shopped Nordstrom). But the zip code data was invaluable for understanding the competitive landscape, and the impact the competitive landscape had on in-store purchases.

Turns out that competition isn't always a bad thing!

Contact me (kevinh@minethatdata.com) for your own, customized project.

Online / Retail Dynamics: Store Distance

A quick note for you today ...

Store Distance plays a major role in how customers "convert", and in how customers use your online channel.

In most of my projects, there are three key store-distance bands.

1. 0-5 Miles:  Here, customers frequently use the website to plan an in-store purchase. Conversions are hard to come by, via traditional metrics, and yet, the website has never been more important. Email, by the way, becomes a major in-store driver in this case.
2. 6-25 Miles:  This band is driven by online research that yields a 50/50 mix of online purchases and in-store purchases. This is the classic "omnichannel" customer that you read about. These customers frequently "do everything". Email is a tool that often drives the customer online, in this mileage band.
3. 26+ Miles:  Even more interesting is the 100+ mile band from a store. Here, the catalog plays a major role in relationship building. These customers tend to buy online after receiving catalogs, at rates higher than for any other segment.

Plug store distance into your segmentation strategy, and then carefully analyze the merchandise customers purchase, the advertising channels the customers purchase from, and the in-store / online mix by mileage band. You're going to find many unique segments that can be actively mined for increased profit.

Contact me (kevinh@minethatdata.com) for your own, customized project.

Online / Retail Dynamics: Merchandise

Yesterday, I talked about how email marketing dynamics are poorly understood in the typical online/retail setting. Similarly, merchandise preference is highly misunderstood. This, too, happens for a good reason - our web analytics platforms are calibrated to explain the impact of advertising on conversion - they are not calibrated to understand the impact of merchandise assortment on customer relationships.

Here's our map, once again.

Look at the three segments on the right-hand side of the image.

The merchandise categories are largely the same.

The customer, by and large, prefers the in-store experience.

Catalogs are a common customer conversion channel among these customers.

What we present to a customer plays a major role in determining what the customer purchases. Catalogers know this best - they've painted themselves into a corner by featuring merchandise that customers age 60+ tend to love.

I once had a client who sent a million catalogs a month, and had a million unique website visitors each month. This client obsessed about the merchandise they put in the catalog - then gave nearly no attention to the merchandise offered on the home page, and on key landing pages. Considering that many catalog recipients never bother to open the catalog, you'd think that an actual visitor on the website, looking at actual merchandise, would be considered a top priority. Not so.

Our segmentation map shows us that different customers have different merchandise preferences. Online, we should feature the merchandise that customers want to see - this isn't a difficult proposition in 2014. And we should realize that what we show the customer dictates what the customer purchases. We unknowingly influence how the customer will behave, often without much thought to the process. In this instance, the catalog aligns with a series of merchandise categories that this brand decides to actively feature - causing the customer to buy those items.

Contact me (kevinh@minethatdata.com) for your own, customized Online / Retail Dynamics project.

Online / Retail Dynamics: Email Marketing

The three most valuable segments are the upper-middle segment, the upper-right segment, and the middle-right segment.

What is the advertising channel that is common across each segment?

Email!

Email marketing is, in my opinion, the least understood of the advertising channels I work with. This happens, quite honestly, because the metrics associated with email marketing success (opens / clicks / conversions) are not aligned with the metrics that really matter (demand, profit).

For this business, there are common characteristics for email buyers.

• More likely to browse than to buy (9.6 to 1, 6.5 to 1, and 5.3 to 1 visit/purchase odds).
• Way more likely to visit the website than the average customer (54.45 visits, 73.57 visits, 30.59 visits, on an annual basis).
• Likely to combine email marketing with another advertising channel.

This analysis brings two thoughts to mind.

1. More than half of the purchases, among the email centric customer segments, happen in-store. This means that the standard open/click/conversion framework completely misses these purchases, and as a result, seriously miscalculates the effectiveness of email marketing.
2. Email buyers tend to be browsers - they visit the site all the time - and therefore, do not need to be given "hard-sell" tactics. These customers, on average, are visiting the website once a week.

I see this outcome, often, among customers who buy from email campaigns. They visit frequently without buying. Conversion isn't the problem here - no - the problem is often a lack of a new message.

Your best customers visit the website all the time. Why bore them with website updates that happen monthly? Or every-other-week?

Some customer segments require a constant supply of new content. Segment your customers, and then give the customers who require new content a steady diet of new content! Stop boring customers!

Contact me (kevinh@minethatdata.com) for your own, customized Online / Retail Dynamics analysis.

Online / Retail Dynamics: High Price Points

Take a look at the upper-left segment in the image below.

Here are some of the purchase dynamics.

• 1.67 Online Orders Last Year.
• 0.18 Retail Purchases Last Year.
• 1.85 Purchases, Total.
• 15.59 Visits Last Year.
• 8.3 to 1 Visit-To-Purchase Ratio.
• Average Price Point Of Items Purchased Last Year = $93.

Now look at the advertising channels this customer uses.

• Affiliates, Natural Search, Paid Search, Referrals.

We have a customer who primarily shops online, who visits 8 times for every purchase s/he makes, who uses all sorts of shopping comparison advertising channels, who uses affiliates to get deals, and then purchases the most expensive items of any customer segment in the analysis.

This may be a segment where the omnichannel playbook is valid. In fact, when you look across the top row of the table, these are all customers who fully utilize the omnichannel playbook. In this case, we have a segment of customers who are clearly looking to save money on the most expensive items the brand has to offer.

Also notice that the merchandise categories purchased by the customer are different than are the merchandise categories purchased by other segments.

This is a customer segment that, in many ways, is different from the rest of the business. This is a marketing-driven customer segment, not an organic brand-centric segment who loves the general merchandise assortment.

Recognize this customer. Realize that this customer is going to visit the website, often (8 visits to 1 purchase). Find ways to communicate to this customer that your pricing is fair.

Contact me (kevinh@minethatdata.com) for your own, customized Online / Retail Dynamics project.

Online / Retail Dynamics: The Catalog Customer

Even in a retail environment, there are catalog customers who behave differently than other customers.

Look at the bottom right segment.

• 0.55 Direct (online) Orders.
• 4.11 Retail Orders.
• 4.66 Total Orders.
• 9.80 Website Visits.
• 2 to 1 Visit-To-Purchase Ratio.

The ad channel this customer uses is catalogs.

Notice that this customer has an 8-to-1 retail purchase skew.

Notice that this customer does not visit the website, less than once a month, to be honest.

In other words, this customer is very different than the average customer ... this customer actively uses the catalog (not the website) as the primary shopping tool, and uses the catalog to buy in stores, not online.

In an omnichannel world, you're told you have to do everything. Wrong. You have to do what is right for specific customer segments. For this customer segment, the catalog is an important component of the purchase process - more important than the website.

Segment your customers, understand their Online / Retail Dynamics, then act appropriately!

Contact me (kevinh@minethatdata.com) for your own, customized Online / Retail Dynamics project.

Online / Retail Dynamics: The Best Customers

Let's look at the upper-right cell in this image.

Wow, what a customer segment. Look at last year's purchase/visit activity.

• 4.71 Direct Orders (Online) Last Year.
• 6.67 Retail Orders Last Year.
• 11.38 Total Orders Last Year.
• 73.57 Website Visits Last Year.
• 6.5 to 1 Visit-To-Purchase Ratio.

This is the most valuable of the nine customer segments in my analysis - these customers spent $1,963 last year. This customer is truly "omnichannel", even as the pundits describe omnichannel behavior. Look at the characteristics ... the customer buys via catalogs, affiliates, email, and social media. My goodness! And the customer splits orders between online and stores, wow!

This is what omnichannel customer behavior actually looks like. The customer essentially "does everything". In fact, if you look at "best customers", you will almost always observe that the customer "does everything", because that's what best customers do!

But most important, this customer visits the website 6.5 times for every purchase.

In other words, stop demanding that this customer buys RIGHT NOW. Stop! This customer visits the website once every six days. That's nuts!! Heck, this customer segment is responsible for 35% of the website visits in the past year from customers who purchased last year.

We need to make a clear distinction among customers - "segmenting them", if you will. Some customers must be converted, now, or you won't see the customer ever again. And then we have loyal customers - why hawk these customers with promotional nonsense when the customer visits the website once every six days?

Contact me for your own, customized Online / Retail Dynamics project (kevinh@minethatdata.com).

Online / Retail Dynamics - Nordstrom

When I worked at Nordstrom, circa 2006, we knew that our "multi-channel" customers exhibited a consistent behavior:

• 3 Website Visits Per Month.
• 2 In-Store Visits Per Month.
• 1 Purchase Per Month, 85% In-Store.

We also knew that online customers, after a first purchase, were more likely to migrate to retail than to stay within the online channel.

We also knew that retail customers, after a first purchase, were going to stay within the retail channel, using the website to research an upcoming in-store experience.

Tactically, you change your thinking when you understand the role that your website and mobile experience play in the customer relationship.

Specifically, there is no reason to demand a "conversion" on a website when the customer uses the website to research an in-store purchase. I know, I know, this runs counter to everything you've been taught. But come on!

In a retail environment (same thing with a catalog environment), the website is far more likely to play a research role than a commerce role. This means you want to recognize the customer upon a visit. If that customer is a "retail researcher", then give the customer what s/he wants - don't demand a purchase today.

Tomorrow, we'll look at one of the segments from yesterday's image (shown below).

Online / Retail Dynamics

During the next two weeks, we're going to talk a lot about what I call "Online / Retail Dynamics" (contact me via email ... kevinh@minethatdata.com ... for your own, customized project).

During the past two years, I've learned that we have a good understanding of online behavior within the context of a visit. Yup, we can clearly see what drove the customer to the website, and we can clearly see what caused the customer to "abandon" the visit. All good, no doubt.

But we're missing the context of a customer relationship, aren't we?

Study the image in this post. There are nine customer segments here, a common outcome in the Online / Retail Dynamics projects I've worked on during the past two years. In the upcoming days, we're going to dig into the nine segments, and in the process, we're going to learn how customers behave in an Online / Retail environment.