October 29, 2015

What Is A Typical Annual Repurchase Rate For An E-Commerce And/Or Catalog Brand?

Quiz: For an average e-commerce and/or catalog-centric business, what percentage of last year's buyer will purchase again this year?

(a) 30%.

(b) 70%.

(c) 88%.

The answer, of course, is (a). Sure, Amazon retains more than 90% of last year's buyers, but they are Amazon, and you are not Amazon.

When you retain fewer than 40 of 100 buyers from last year, you are in Acquisition Mode. The focus of your business must, and I repeat must be focused on finding as many new customers as possible, at as low a cost as possible.

Of course, everything you read says the exact opposite. Trade journals, consultants, conferences, gurus, vendors, they all make money when you do the opposite. They want you to spend money to encourage customers to become more loyal, even though the odds of the customer becoming "more loyal" are maybe one in ten, if you are lucky.

I've worked with more than 200 brands since founding my consultancy way back in 2007. The most successful clients I've worked with have low-cost customer acquisition programs that continually fuel the growth of the business at a profitable rate. The least successful clients I've worked with spend money with ever vendor imaginable in an effort to manufacture growth.

October 28, 2015

Yes, Prioritize Customer Acquisition Over Customer Loyalty!

Can I show you something?

For a business, I built a model that predicted the probability of repurchasing in the next twelve months. If a customer possesses a 60% or greater chance of repurchasing in the next twelve months, the customer is deemed "loyal".

Quick Quiz: For this business, what percentage of customers who purchased in the past month are deemed "loyal"?

(a) 18%.

(b) 33%.

(c) 79%.

For the business I am analyzing, the answer is (a).

By recency (months since last purchase), here is the percentage of customers who the model deemed to be "loyal".
  • Recency = 01:    17.55%
  • Recency = 02:    15.57%
  • Recency = 03:      9.05%
  • Recency = 04:      9.00%
  • Recency = 05:      9.66%
  • Recency = 06:      8.31%
  • Recency = 07:      6.91%
  • Recency = 08:      6.24%
  • Recency = 09:      4.88%
  • Recency = 10:      3.85%
  • Recency = 11:      3.57%
  • Recency = 12:      2.92%
  • Recency = 24:      0.40%.
  • Recency = 36:      0.12%.
  • Recency = 48:      0.00%.
This is a typical catalog / e-commerce outcome, for a typical catalog / e-commerce business with a 30% annual repurchase rate among twelve-month buyers.

If you run simulations, as I've been running for the past twenty-plus years, you very quickly learn that it is terribly hard to push customers in this type of business into "loyal" status. What are you going to do? Be honest! Offer points? Free shipping? 20% off. Gift with purchase? You're already doing all of that nonsense and if anything, annual repurchase rates have declined over time, right? And everybody else is already doing all of that nonsense. Wait. Apple doesn't do any of that stuff. Oh, that's right, they have great products. Never mind.

So you and I already know and understand that you're never going to build a crop of "loyal" buyers that is large enough to "push the peanut", as a former CEO used to say to me.

But new customers?

We can always find new customers. We can always grow your customer base by finding new customers. And when you increase the number of new customers by 50% for a few years, we will almost surely increase the number of loyal buyers by 50% in several years.

We can waste money forcing a customer into loyal status - which isn't a loyal status in the first place because the customer didn't get to loyalty via love of merchandise but instead via games you played to force the customer to get there, costing you profit along the way. Or worse, the customer would have gotten there but we accomplished the same outcome while giving away points and free shipping and 20% off. Woo-hoo!

Or, we can waste money finding new customers - knowing that in time, some of the new customers will become loyal without any games or discounts or promotions whatsoever.

If I were to ask 100 of you how you felt about this, 95 of you would tell me I'm crazy. I know this, because I've watched your unsub rates from email over the past week and I've read your comments. I haven't been given any actual customer data that supports the loyalty hypothesis, but I've been given criticism.

And yet, when annual repurchase rates are under 40%, business success is almost entirely determined by a combination of merchandise productivity and low-cost / high-volume customer acquisition programs. Run the simulations to understand the dynamic for yourself.

Keep in mind that 80% or more of the catalogers / e-commerce brands I've worked with possess annual repurchase rates under 40%.

So that's what the data tell us. We really need to adjust our mindset. We need to focus our energies on finding low-customer new customers. Soon!

What Happens When New Stores Open?

Quiz: When you open a new store in a new market, what happens to e-commerce sales in that market?

(a) E-commerce sales increase, because in an omnichannel world, customers simply spend more, often up to 9x more, when you give them more channels.

(b) E-commerce sales decrease as sales migrate out of the website to a new store, then resume a normal trajectory.

Well, if (a) were true, then retailers would be opening stores all over the place, because new stores would generate incremental online sales and would generate in-store sales and the math would be multiplicative and you couldn't help yourself but to double your store portfolio.

But that doesn't happen, does it? If it did, the biggest retailers wouldn't be closing stores all over the place.

Retail is a big game of whack-a-mole. When you open a store, you cannibalize e-commerce sales, you cannibalize sales from other stores in the market, and you grow in-store sales in the geographic ring where the new store is opened. The math is almost never multiplicative, as the pundits suggest it is. Cannibalization almost always happens. The smartest businesses understand cannibalization, and they account for it.

October 27, 2015

How Do You Determine If A Customer Is "Loyal"?

This one has always been an easy one for me. Now, I get it, you're not going to agree with me, and that's fine. But when you don't agree with me and you don't have your own definition of what a loyal customer is, well, you've got a problem, don't you?

In other words, you should have a one sentence description of what a loyal buyer is.

Here's my one sentence description of what a loyal buyer is, in retail / e-commerce / cataloging:
  • "A loyal buyer has a 60% or greater chance of purchasing again in the next twelve months."
When I worked at Nordstrom, the marketing team used this definition:
  • "A loyal buyer spent at least $750 in the past twelve months."
Those are two very different definitions, don't you think?
  • My definition is forward-looking.
  • Their definition was backward-looking.
Neither definition is right/wrong.

The forward-looking definition gives me a lot of flexibility. For instance, these two customers may both have a 60% chance of buying again in the next year.
  • 1 Day of Recency, 1 Life-To-Date Purchase.
  • 33 Months of Recency, 22 Life-To-Date Purchases.
Maybe I just purchased a table from Wayfair and am thrilled with my purchase! I am as loyal today as I'll ever be, but the half-life of my purchase is short, and if something magical does not happen in a month, I'll forget my one purchase.

Maybe I purchased from Macy's 22 times, but their omnichannel efforts have not inspired me for nearly three years. I'm still loyal, but I am waiting for Macy's to do something special.

Measure the inflow and outflow of "loyal" buyers in your buyer file. Track it monthly. Understand if you are doing the things that need to be done to generate a loyal buyer file.

October 26, 2015

But That Path To Loyalty, I Can Greatly Influence It With A Postcard Or Retargeting Effort, Right?

Here's the challenge most of our businesses face.
  • 12 Month Probability of a 1st to 2nd Purchase = 30%.
  • 12 Month Probability of a 2nd to 3rd Purchase = 50%.
  • 12 Month Probability of a 3rd to 4th Purchase = 60%.
Yup, in 7 of 10 cases, this is what it looks like. It typically takes four purchases before a customer can even be thought of as being "loyal". Now, I get it, some customers purchase after 37 months blah blah blah ... but for the sake of argument, let's just multiply out the numbers above for 1,000 sample first-time buyers:
  • 1,000 * 0.30 * 0.50 * 0.60 = 90.
There you go.

You acquire 1,000 customers, and 90 become "loyal". Throw in all of those lapsed buyers who eventually purchase, and maybe you get 150 buyers to become "loyal" ... eventually ... and it takes three years to get there.

Three years!

And you only get 150 of 1,000 customers to a state of "loyalty".

Now let's pretend that you have some sort of amazing postcard or email or campaign-centric program that you beat your customers over the head with on a weekly basis, incremental to everything else you are doing. You might bump that 150 number by 5% ... another 7.5 customers.

Do you see why most loyalty efforts are a colossal waste of time?

Why not spend the same amount of energy and money finding a new customer instead?


October 25, 2015

Pro Humans


It's fun to have a research brand "imagine a future" where humans earning $12 an hour are fired, replaced by automation that is ultimately sold by the brands that pay the research organizations to write articles about imagining a future where technology replaces humans. It's fun to provide thought leadership!! In fact, it's fun to not have to be accountable for firing humans. Ask your favorite researcher who writes thought leadership articles like the one referenced above.
  • Aside: Years ago, the salary structure of my team was being modified by a compensation manager. The individual modifying the salary structure certainly enjoyed the changes - heck, he was bonused for exceeding his objectives in making the changes. I asked the compensation manager to help me out, since half of my team would take a pay cut as a result. The compensation manager said "no". I then refused to implement his changes unless he came to my department meeting and communicated the changes to my team, taking accountability for the changes he was forcing upon my analysts. His response? "I don't want to do that!" Well, of course you don't want to do that! It's fun to cut pay for people you don't know and don't ever have to talk to. Regardless, I refused to implement his changes unless he addressed my team. Needless to say, my team was promptly compensated. It's funny what happens when the individual responsible for ruining lives has to face the accountability of communicating said ruination.

The whole omnichannel movement is offensive. You know it. I know it. Vegetable lasagna knows it (bonus points to those who know what that reference means). Vendors demand that you fire humans and install their technology instead. Consultants claim that "the customer", whoever that is, would rather be served by a robot than a human. They tell you that you have to align all your channels or you are "dead", even though Amazon aligns nothing and crushes your online business in the process.

Try to be a human being.

Try to have compassion for others.

You can transition your business to the future and maintain a semblance of humanity in the process. Stay away from the cold attitude of the linked article. Be kind.

In The Old-School World, This Was Called Your "Hotline" Program


To their credit, these folks have realized that the half-life of digital data is about eight seconds, so they're taking advantage of that via print.

We had this ability at Nordstrom in 2006 ... half-life prevented us (and the vast majority of marketers) from taking reasonable action ... there was simply no reason to spend money to hotline anything based on online activity when you could personalize digital information (especially email) in near real-time. Why send a postcard that arrives in days when you can send something digital right now? And that was in 2006.

But for a lapsed catalog-centric buyer with < 10% annual repurchase rate and an average age > 55, have at it!

October 22, 2015

Managing Conversion Rates

Question: You make a change to your website. You used to have 100,000 visits and a 3.0% conversion rate. Now you have 93,750 visits and a 3.2% conversion rate. Average order value is $100 in both cases. Did you fundamentally improve the business by improving conversion rate?
  1. No.
  2. Yes.
Well, both answers get us to 3,000 orders at a $100 average order value.

In other words, pre-change and post-change, we generate $300,000.

So, no (1), we did not fundamentally improve the business by improving conversion rates.

This is a common dilemma out there in the real world. We make changes to "reduce friction" ... and in this case, we reduced friction. But we didn't increase sales. And if we can't increase sales, well, then the effort is largely wasted.

In other words, if the customer used to visit six times before buying something, and now visits five times before buying something (and spends the exact same amount), we reduced friction but we did not improve sales, and we did not improve profitability, so maybe we need to spend time elsewhere (or change how we measure success - go talk to Google about that one).

October 21, 2015

Make A Profitable Choice

Question:  You mail a monthly 128 page catalog. You decide to test a 64 page catalogs against the 128 page standard offering. You learn that the 128 page catalog generates $4.00 per book, yielding $0.40 profit per customer. You learn that the 64 page catalog generates $3.40 per book (85% of total), yielding $0.54 profit per customer. What do you decide?

  1. You decide to continue mailing the 128 page catalog because your printer tells you that it generated more demand than the 64 page catalog.
  2. You mail the 128 page catalog to a circulation depth that yields a break-even result. Then, you mail the 64 page catalog to the remainder of the circulation, as long as the 64 page catalog can be mailed profitably.
  3. For every customer, you run the math, and if 128 pages is more profitable, send 128 pages. If 64 pages is more profitable, send 64 pages.
You already know the answer, don't you?


Here's what I observe.

  1. 50% of catalogers would go with (1).
  2. 35% of catalogers would go with (2).
  3. 15% of catalogers would go with (3).
But, easily, the most profitable answer is (3). Bigger companies have figured this out, and they make a ton of profit as a consequence. 

October 20, 2015

Lifetime Value

Question:  You lose $8 profit acquiring a new customer. The customer generates $7 profit in year one, $5 profit in year two, and $3 profit in year three. Which time horizon should you maximize long-term value upon?

  1. You lost $8 profit, so stop marketing to this source of new customers.
  2. You lost $8 profit, and you do not make up the difference in year one, so stop marketing to this source of customers.
  3. The customer generates $15 profit in three years, more than offsetting the $8 lost in year one. This customer source is profitable, and should be profitably mined.
  4. You were not give enough information to make an informed decision.
What is your answer?

The right answer is to run a long-term simulation, and let the simulation determine if you generate enough long-term profit to maximize shareholder value. In other words, the answer is (4).

On average, simulations show that there is a direct correlation between annual repurchase rates and the length of time you are willing to wait for payback.

In other words, if a new customer has a 20% annual repurchase rate, then there is very little future value generated, and as a result, you're going to have to break-even up-front on customer acquisition costs.

If a new customer has a 50% annual repurchase rate, then there is significant future value generated, and as a result, you may be able to wait three years or four years to make up acquisition costs.

But you have to run the simulations to know the right answer.

October 19, 2015

Paid Search Clicks and Email Marketing

Today's Question:

Question:  If you execute an email mail/holdout test, and you learn that 20% of paid searches stop happening if email marketing is discontinued, what should you do?

  1. Spend more on paid search because paid search helps make email marketing look better.
  2. Attribute 20% of all paid search demand to email marketing.
  3. Measure all paid search demand and paid search clicks and paid search costs associated with the clicks back to your email marketing program, based on the results of the mail/holdout test.
The answer is an easy one, isn't it?

The answer is (3).

And if I ask 100 readers how many actually execute (3) above, I'll find that fewer than 5% do the right thing.

Your attribution work must attribute demand, clicks, and costs back to email marketing, should you learn that email marketing causes customers to perform more paid search clicks.

October 18, 2015

Loyalty Programs

A quiz question for you:

Question: On average, where do loyalty programs pay off the most?
  1. Among the most loyal customers (i.e. those with 60%+ annual repurchase rates).
  2. Among average customers (i.e. those with 40% - 60% annual repurchase rates).
  3. Among marginal customers (i.e those with 0% - 40% annual repurchase rates), because the loyalty program pushes the customer from inactive to loyal status.
The answer?

Across all of the loyalty work I've done, over twenty years, it is obvious that loyalty programs work when customers are already very loyal.

Here's the problem. Loyalty programs have very few levers - points leading to discounts, or discounts, or promotions, or freebies, or service enhancements, or combinations of the above. Those tactics must generate profit, in order for a loyalty program to succeed. Just as important, those tactics must increase sales significantly, or what is the point of a loyalty program outside of data capture?

Now, if a loyalty program increases customer-level demand by 10% (it's often between 5% and 10%, measured incrementally), then we have two situations.
  1. A customer with a 30% annual repurchase rate and 1.5 purchases per year nets out at 0.45 purchases per year. The loyalty program increases annual purchases per year from 0.45 to 0.45*1.1 = 0.495 purchases per year. That doesn't change anything.
  2. A customer with a 70% annual repurchase rate and 6.0 purchases per year nets out at 4.20 purchases per year. The loyalty program increases annual purchases per year from 4.20 to 4.20*1.1 = 4.62 purchases per year.
In a low retention environment, the loyalty program is accountable for 0.045 incremental purchases per customer per year.

In a high retention environment, the loyalty program is accountable for 0.42 incremental purchases per customer per year.

The answer above is (1). If your annual repurchase rate (across all twelve-month buyers) is below 60%, your loyalty program initiatives are not likely to move the needle much.

October 15, 2015

On Friday, It Begins!!

Yup - as of the end of day yesterday, I received all of the data from the lucky participants in the MineThatData Elite program. Shortly, this audience of pioneers will be rewarded with knowledge of their own merchandise productivity, as well as an "average of averages" of merchandise productivity across all participants.

Ok, so you elected to not participate this time. But by now, you're saying to yourself, "Why wouldn't I want to know how my merchandise productivity stacks up against other participants? Is my business sluggish because of what I'm doing, or because of larger macro-economic issues?" You're wishing you got involved sooner, when the cost was locked in at $1,000 per run.

Well, after this run, there's going to be a lot of knowledgeable folks who have a competitive advantage over you. You'll want to join them. And you will join them ... you'll lock in at the $2,000 rate before February 1 (at which time, rates increase to $2,500 per run).

Get necessary approvals now - budget $6,000 for the three runs I anticipate in 2016. Then, you (too) can be part of the magic!

October 14, 2015

Price Across Channels

You are sometimes told that you must create a seamless / frictionless shopping experience across channels, right?

And yet, here's Amazon, the very company that is putting you out of business, offering the same text at a veritable plethora of prices across digital or dead-tree channels. Heck, there are differences among paperbacks. And look at all the profit that the publisher gets when they sell a unit via Kindle, at essentially no physical cost whatsoever.

But, again, the experts tell you that you must offer a seamless / frictionless shopping experience across channels - telling you that the customer "demands" that you offer the same item at the same price in all channels. Why does the customer demand that from you but not demand that from a company (Amazon) that the customer clearly prefers?

Find out how much a customer wants to pay for the merchandise you sell. If you must maintain pricing parity across channels, then why not take comparable items and vary prices across the board on comparable items - for a month or two? What would stop you from learning the truth about customer response?

October 13, 2015

Optimize This!

You visit a website ... and it is nothing but ads ... nothing! Ads on top of ads. Pre-rolls on top of the content you're hoping to see.

I know, I know, you're going to yell at me ... you are going to tell me that I'm an idiot who doesn't realize that somebody has to pay for the content I wish to view.

But that's not my point.

My point is that there is no content here. None.

In calculus, this is called a "local maxima". It's a case where you optimize yourself to a peak ... and if you go in any direction, you drop from the peak, so you just stay here ... not realizing that if you drop from the peak, you may end up at a higher peak somewhere else.

If you don't know calculus, and you took your "Master Big Data And Optimization In 30 Minutes" course, you may not realize that you are stuck at a local maxima. And consequently, your website is nothing but ads.

October 12, 2015

But What About The 25% Of Catalogers Who Have Young Customer Bases?

Young (in catalog terms) meaning age 35-54, of course.

There are things that these businesses do that are fundamentally different than what most catalogers do. Some of this is "chicken 'n egg" logic, and cannot be unraveled, of course.

*** Their dependence upon the co-ops has always been less than the typical cataloger, thereby seeding their business with younger names who like merchandise tailored to younger customers.

*** They sell more "fashion oriented" items.

*** Their merchandise aligns with stuff needed in a new home, or aligns with stuff needed by families with children. In this way, they draw the younger portion of the co-op algorithm (where there is less competition, interestingly), and they are far more appealing to the typical individual prowling out there for the best deal on Google.

*** They offer free shipping more often than discounts/promotions.

*** Their online marketing departments are fully staffed, and are excellent.

*** They work at a pace/urgency that most catalogers cannot match.

*** Their attribution programs equalize between online marketing and catalog marketing. Traditional catalog marketers with older customer bases assign online orders back to the catalog without questioning the results.

*** They test, often.

*** They have smaller page counts.

*** They seldom mail remails, and in the process, they do not bore their customers as much.

*** They shift their customers from online to the mobile web, and then into apps. The most loyal customers end up in apps, not in catalogs. The opposite is true for traditional catalogers.

*** In "digital" and "social", they are far more likely to tell a story, without worrying about return on investment. Traditional catalogers tell stories in print, then align digital with print, causing digital to always lag far behind the customer, and then demand that everything is measured and attributed back to the catalog.

*** Email marketing supports merchandise, is merchandised dynamically and/or customers are targeted via one of the five best creative/merchandise treatment combinations several times per week. Traditional catalogers use email to support the catalog, and have only one version of an email campaign that is sent to everybody.

*** The online marketing team has minimal catalog experience, and that is not viewed as a negative.

*** Decisions are made in a nimble manner (as vendors like to say, these folks are "agile"). They will change things online without catalog alignment, capitalizing on sales opportunities. Traditional catalogers wait to change until the catalog can "catch up", thereby fully integrating channels.

*** Planning is centralized around merchandise and events. Traditional catalogers centralize planning around catalogs.

*** They often have a retail channel, where younger customers by default are likely to shop. This drives the merchandise assortment much younger, spilling over into e-commerce and cataloging.

Obviously, these are generalizations of things that I observe. I've witnessed catalogers with 40 year old customers and the oldest-school techniques you'll ever see. I've also seen catalogers with 70 year old customers who execute programs that mobile-savvy brands would envy. 

But you get the picture. I share this with you, in hopes that you'll think about what it all means.

October 11, 2015

Urban Outfitters

You've heard about this one, undoubtedly (click here).

Twenty five years ago at Lands' End, we were asked to pick/pack/ship orders in early December ... no, not on the weekend ... but on a non-descript Tuesday. I distinctly recall picking the order, setting the order aside, packing the order form, zipping the envelope, and sending the sewn package down the line. Then I looked in front of me ... three items sat there, lonely. Somewhere in Minnesota, a perfectly humble Protestant family received a package containing only an order form / receipt. Merry Christmas from Lands' End!! Moments later, I was relieved of my duties ... not as a statistical analyst ... but as a distribution center employee.

During a seventeen inch snowstorm, I committed myself to shoveling a whole road so that I could get to work. I got to work, and was promptly re-directed to one of our call centers. I took two orders ... TWO ... and messed them up so bad that I was relieved of my duties ... not as a statistical analyst ... but as a call center employee. I was reassigned to the switchboard, where my skills were more than adequate to handle the complete lack of incoming corporate call traffic on a snow day.

At Nordstrom, we were required to work sale events. Yup, there's Kevin, working in a women's apparel department at Northgate Mall, stocking shelves, taking abandoned merchandise out of dressing rooms, and answering questions I was completely unqualified to address ("why does this fit different in a size eight from this item that is a size six, when it's the same brand?"). I hated every minute of this work, work that happened three times a year for six consecutive years.

At Avenue A, it was expected that employees worked eighty hour workweeks. I recall the twenty-three year olds showing up at 9:30am, going home at 9:30pm, playing ping pong for three hours at a time. I worked 6:30am - 5:00pm. I recall being asked why I wasn't "committed to the mission?" I recall asking how a twelve hour workday that included three hours of ping pong helped increase shareholder value when I was working straight through a shorter day? No answer to that response, of course.

At the Garst Seed Company, we were required to work additional hours on weekends in October to handle "harvest season". We'd receive a bumper crop of data that could not be analyzed in the requisite fifty-five hour work week, so Saturday and Sunday were required work days in October. I recall having a weekend trip planned - I did not show up for required weekend work, and it didn't help my career one bit ... "what is wrong with you, don't you know that this is 'harvest' and you have to work weekends?"

Salaried employee are required to donate time for free. Show of hands - how many of you worked just forty hours last week? I get emails from east coast folks time-stamped at 10:30pm. I get emails from west coast folks at 5:30am. My team at Nordstrom FTP'd files at 2am on a Sunday Morning - a day of rest, no less. 

There is no rest for a salaried marketing employee.

There is nothing uncommon about what Urban Outfitters asked salaried employees to do. Sure, it appears distasteful. Sure, it might be a way for the finance team to celebrate yet another cost-cutting measure. But most salaried employees are already being abused through the magic of a sixty-hour workweek and non-stop on-call availability. 

The question isn't whether what Urban Outfitters is doing is right/wrong. The question is how did we get to a point where salaried employees are expected to work sixty or more hours a week, seven days a week no less - essentially not being paid for a third of the work they already do? It's harder to answer that question. It's easier to point to Urban Outfitters instead.

Look at your own career. How many hours north of forty do you feel compelled to volunteer, for free?

October 08, 2015

How Do I Know If I Have A Merchandise/Age Interaction Problem?

This one is simple, folks.


Go run your comp segment analysis, segmented by age of customer. Look at 2x buyers, and if you are a cataloger, segment by customers age 18-49, customers age 50-64, and customers age 65+. If you don't have enough customers age 18-49 to run the analysis, you already know your answer.


But if you do, and you see this (on an annual basis), then you have a merchandise/age interaction problem.
  • 18-49 Customers:  This Year Comp = -10%. Last Year Comp = -9%.
  • 50-64 Customers:  This Year Comp = -3%. Last Year Comp = -4.
  • 65+ Customers: This Year Comp = +7%. Last Year Comp = +4%.
  • Overall 2x File: This Year Comp = +1%. Last Year Comp = -1%.
This is the signature. Older customers love your merchandise. Customers under the age of fifty are running (not walking) away from the business.

On the surface, the business is fine ... flat comps over two years.

But under the covers, younger customers do not like the assortment, while older customers crave the assortment.

Run the analysis. It isn't hard to run a comp segment analysis. Seriously. Go do it!

And if you don't have the resources to run the analysis, contact me (kevinh@minethatdata.com). Contact me soon, too. You want the results before you get too far into 2016 to do anything about it, right?

October 07, 2015

How Algorithms Are Evolving Our Businesses

For so many catalogers, the issue in 2015 is new customer acquisition. It's become terribly hard to acquire new customers.

There's a good reason for this. 

Algorithms.

Algorithms need data.

The smartest companies get you to pay them to collect data. Think Google - you pay them $0.60 a click, and in the process, they get to collect data about the person who clicked. Oh, and they get to collect data about every non-click as well. You fund the non-clicks. So you pay Google money, Google sends you clicks, and then Google asks you to use their software to analyze whether the clicks you paid for purchased or not. When you use Google Analytics, you give Google more data about how your customers behave, data they would not normally receive. All of this data is algorithmically used by Google to benefit Google.

As a result, Google knows that the catalog customer is 62 years old.

You pay Facebook, too. You were told to build your audience on Facebook, organically, for free. Then, when your audience peaked, Facebook figured out how to get you to pay to speak to the audience you helped Facebook create. Nicely done. Facebook takes your money, and slices and dices your customer based on behavior within Facebook and the sliced/diced data you indirectly pay some of the co-ops for ... yes, you give your data to some of the co-ops for free, they make you pay for sliced/diced names, then they partner with Facebook to create even more thorough profiles of customer behavior ... offline, online, and within Facebook. All of this data is algorithmically used by Facebook & some of your co-ops to benefit Facebook and some of your co-ops.

You pay Amazon, too. You sell your products there, allowing Amazon to collect a fee on each purchase, and more importantly, collect data on how your customers behave when shopping for your products on Amazon. This allows Amazon to craft algorithms to target your customers.

You pay the retargeting folks ... those folks track your customers all over the web, algorithmically determining what your customer should see as your customer travels the wide expanses of the online ecosystem.

Each of the 000s of companies you are paying are building algorithms, using algorithms, or are letting machines learn how to build their own algorithms.

The algorithms interact with each other, creating unexpected outcomes.

Catalogers know this all too well.

In the past five years, the average age of the catalog shopper has gotten old, quickly. The algorithms are interacting with each other. They know that the catalog shopper averages 62 years old, often older. The traffic Google and Facebook and the co-ops and the retargeters and affiliates and countless others send you represent traffic that peculated to the top of the algorithm flow chart. Those customers are older. Older customers have specific merchandise preferences.

Now your own algorithms take over, and further impact you. Sure, you call it "reporting", but be honest, it's just a simple algorithm. The co-ops and Google and Facebook and the retargeters and affiliates and countless others send you older traffic, and the older traffic has specific merchandise preferences. In 2000, those were the preferences of a 47 year old. That customer cared about mainstream, middle-aged merchandise. Your merchants responded, offering additional mainstream, middle-aged items that mainstream, middle-aged customers wanted. But today, your merchants offer stuff that older customers like, and your merchants offer stuff that younger customers like. The 62 year old sent to you by the vendor community sees each assortment, and buys the stuff that older customers like. Your algorithm (reporting analyzed by your merchants) clearly shows that stuff that 62 year olds like is the stuff that is selling best. Your merchants respond by getting more of that stuff - the opposite of the "fast fashion" movement that is fueled by a mobile vendor ecosystem that interacts with retailers catering to a younger customer.

If you are a catalog that has been around for at least twenty years, do me a favor. Go pull out a catalog from 1995, and compare it to a catalog from 2015. Objectively look at the styles featured in the first twenty pages of each catalog. Tell me that you are still marketing to a 47 year old customer, after you look at the styles offered and the models wearing them.

The algorithms ... complex machine learning algorithms at Google all the way down to merchandising reporting ... are interacting with each other, driving the age of your customer base north, driving the merchandise assortment to something that an AARP member would crave. Your merchandising team buys more of the stuff that an AARP member would like, and as a result, Google and Facebook and the co-ops all respond in kind, sending names that would like this product, driving your merchandise assortment even further from the mainstream.

You won't find many catalog vendors or catalog consultants or catalog gurus who talk about this. They don't want you to know this, because if you knew this, you'd question your long-term viability, and you'll question your circulation plan going forward. When you question your long-term circulation plan, you put vendor livelihoods at risk. They don't want that. So nobody talks about this topic. Everybody keeps quiet. 

And that does a real disservice to our businesses, don't you think?

Talk about this topic. Think. Have an honest discussion. Or not. Either way, the algorithms are sending your business down a path you didn't anticipate five years ago. Take control back from the algorithms. Please!

October 06, 2015

Hiding The Numbers


Then Bed Bath & Beyond honestly tells what is happening (most retailers don't).
  • E-Commerce = +25%.
  • Store Comps = -1%.
If you remember your high school algebra, then you know that the ratio above suggests that e-commerce sales are +/- 6% of total (remember, new stores are not counted in store comps, so the actual answer of 6.5% is overstated).

Why do retailers hide poor store comps in a blended average of store comps and e-commerce growth?

Well, as your e-commerce channel grows, your blended comp grows, as long as e-commerce keeps growing.

When e-commerce is 6.5% of the total, the blended average of +25% and -1% is +0.7%.

When e-commerce is 12% of the total, the blended average is +2.1%.

When e-commerce is 20% of the total, the blended average is +4.2%.

As retailers figure out how to grow e-commerce, store sales decrease ... and as store sales decrease, the reporting changes, skewing to e-commerce, enabling the perception that growth is fantastic.

Regardless, in every situation, store sales are in decline - and eventually, stores are closed - they have to be closed. When a store is closed, it is not included in the comp measurement.

In other words, as e-commerce gobbles up commodity-based items that used to sell in stores, e-commerce drives retail comps further negative while growing e-commerce, causing the business to report positive comps as it closes stores, accelerating the mix to e-commerce further, causing the business to continue to report positive comps as the business contracts.

That's the future. That's what is coming ... e-commerce growth and store closures and business contraction, all hidden with an overall and ever-growing positive comp store sales measure. #Omnichannel!!

October 05, 2015

The End of the Baseball Season

I grew up in Wisconsin, and I live in the Pacific Northwest. This means that I follow two teams.
  • The Milwaukee Brewers.
  • The Seattle Mariners.
By the time we got to May 1, there was little reason to watch either team.

Now think about how each team dealt with their challenges:
  • Milwaukee fired their Manager before April ended. They traded their best player away before the July trading deadline. They restocked their farm system in July and August. They announced that their General Manager was retiring, and hired a new General Manager (30 years old) in September, an individual with an analytics background. They gave extensive playing time to minor league players in September.
  • Seattle waited and waited and waited for something good to happen until late August. They then fired their General Manager a month past the trading deadline, and hired a new General Manager in late September.
Which team got a head-start on the future?

Ok, how about you? It's October. You pretty much know how your year is going to turn out, barring a miracle. If your business is below plan, are you taking the approach Milwaukee took, looking toward the future, or are you being like Seattle ... hanging on, not making changes, hoping for things to change ... not making changes until it is very late in the season?

Discuss.

October 04, 2015

The Data Is Wrong!!

Did you see this article from Friday (click here)?

First, look at the ad next to an article about bad weather data:




Alanis Morissette would agree it is ironic that an article talking about how terrible initial data conditions lead to terrible forecasts is supported by an ad about a conference featuring technology that uses incomplete data to put ads in front of customers that are never clicked on - a technology that you use to bid to pay higher-than-necessary rates to put an ad in front of a customer that seven in ten thousand (i.e. nobody) will respond to.

When your co-op sales rep tells you that they have a new "coherence model" (or whatever name they call the next version of models that are cousins to the models used in 1995) and that they're seeing breakthrough results, you have to question the statement.
  • Is the data being used to create the model "right" in the first place? Ask your co-op sales rep to see the data being used as an initial starting point.
  • Are the models that sit on top of the data "right" in their assumptions?
  • Ask the co-op rep to share the model coefficients with you - it's your model, you are paying for it, so you should get to see what you are paying for, right?
Ask your co-op sales rep to answer each question. If you don't get a valid answer to any of the three questions (and you probably won't), then ask yourself why these folks are still trusted partners? Why would your trusted partner hide the truth from you? (and co-op folks, this works in reverse - if your clients won't share profit/loss data on the names you give them, are these your trusted partners?).

Let me tell you a story. I was asked to visit a company and referee a discussion between vendor statisticians and in-house statisticians. What a mess! The vendor used the same data that the in-house statisticians used. But each party transformed the data in different ways, so their starting points were fundamentally different. Different starting points mean different outcomes (like in the example with hurricane forecasts). And different starting points mean different modeling approaches. Each party utilized fundamentally opposite tactics. The vendor had hundreds of variables in the model, optimizing profit. The in-house statistical team had a half-dozen variables optimizing sales. Again, both tactics are wrong (never use 100s of variables in a model, the last several hundred have no impact - the concept is called 'parsimony' ... and modeling sales instead of profit means that you reward customers who return a lot of merchandise and/or purchase via discounts/promotions), and because both tactics are wrong, both tactics yield bad outcomes. Of course, both tactics are doomed because they start with bad data (i.e. bad variables).

But man, what an argument, where two parties who are wrong yell at each other about how wrong the other side was! It almost sounded like our political process.

If your in-house statistician won't share the details of the model building process with you, find a new in-house statistician.

If your vendor co-op or retargeter won't share the details of the model building process with you, find a new vendor co-op or retargeter.

If your attribution vendor won't share the intimate details of the modeling process used to dictate your advertising investment algorithm, then fire the attribution vendor and find a new one who is honest and transparent.

In weather, we can clearly see when one modeling process, #datadriven and all that, is still woefully wrong.

In marketing, do we know if our trusted analytics partners are right or wrong?

October 01, 2015

Highly Targeted Digital Ads That, Well, Just Read The Article.

Look at the ad that Google decided to interrupt my viewing experience (of the VW scandal) with:



It's a local Lexus Dealership ad featuring a used 2013 Volkswagen!!! That's exactly the car I want to buy when I'm viewing a video about how VW cheated their customers and the rest of us who have to breathe the post-processed air exhausted out of VW cars.

Remember stuff like this the next time the sales person arrives in your office, promising to use digital marketing to optimize the customer experience via the right ad to the right customer at the right time.

Items That Appear In Multi-Item Orders

In a typical Life Stage Analysis within a Merchandise Dynamics project, it is common to see exaggerated trends when comparing first-time buy...