Mining Data

Today, Lauren Fetzer, CEO of Fetzer's Footwear, has invited me to join her in a meeting with her Executive Team. The team is going over plans for 2011. Lauren is in her office, listening to her iPod Touch.

Kevin: "What are you listening to?"

Lauren: "Always The Last To Know by Del Amitri. The lyrics aren't what matters, I just like listening to this before Executive meetings because so often something comes up, and I'm the last to know."

Kevin: "Alright."

Lauren: "Let's go."

We enter the conference room.

Lauren: "Ok folks, let's get going. Penny, what did sales look like yesterday?"

Penny Parker (VP, Marketing): "The website was up 17% yesterday vs. last year, traffic was up 22%, conversion rates were down significantly, continuing a theme we've seen for a long time now. Retail comps were up 5%, led by Downtown Seattle at +22%. Alderwood was -2.8%"

Bart Cox (VP, Stores): "Don't ask, Lauren."

Bill Bledsoe (VP, Logistics): "Something doesn't make sense, folks. We've worked so hard to improve the website experience for our customers. We've implemented a myriad of improvements. Yet, no matter what we do, conversion rates don't improve. Well, they seem to improve for a while, then they just fall back to our baseline."

Ashley Zimmerman (VP, Merchandising): "What's on the docket to improve conversion rates in 2011?"

Connie Simpson (VP, Finance): "And given the potential sales we could generate if we improve conversion rates, it makes perfect sense to spend money to improve conversion rates, so I'm not here to get in your way."

Penny Parker: "Let's pull up a spreadsheet (Penny clicks on her laptop keyboard, a file opens and is displayed on a large plasma monitor). In 2011, we are planning to do a series of website improvements, based on testing we've done during the past year. You can clearly see the laundry list of improvements on the monitor. We have at least two dozen e-mail campaigns that will support our improvements and drive traffic to the website. We strongly believe, based on our tests, that we need to focus more on the product on the home page, and utilize less lifestyle imagery. The tests show we can improve conversion rates by 8.43% if we do this on an ongoing basis."

Ashley Zimmerman: "We've gone over this over and over and over again. We need to communicate the lifestyle aspect of our brand, or we don't have a brand, we're just Zappos without the social media tie-in."

Penny Parker: "And our testing shows that presenting the lifestyle aspects of our brand is counter-productive to driving sales and profit. We have the data to prove it."

Bill Bledsoe: "Are the result statistically significant?"

Penny Parker: "Absolutely. Our tests showed that there is only a 2 in 100 chance that the results are wrong."

Bill Bledsoe: "Wow, that's amazing. How often in business can you count on something ninety-eight times out of a hundred?"

Penny Parker: "I agree! That's why we need to stay the course. We need to focus on what has always worked. I think we just need to make our campaigns work harder. Everybody needs to step up to the plate and be accountable for delivering results."

Ashley Zimmerman: "That works to a point. If you follow that strategy to the extreme, we'd still be selling Colorado boots from the 1980s. At some point, you have to take risks."

Penny Parker: "My team measures risk. We test products against each other, we test creative treatments, we test home page offers, we test it all. When we follow the outcome of our test, we succeed."

Connie Simpson: "So why is it that we do all of this testing and optimization and analysis and nothing ever gets better, Penny? What was our annual retention rate last year, 55%? It's been 55% for a decade. You keep improving everything, and the needle doesn't move a bit."

Penny Parker: "I disagree. Had we not tested and optimized, maybe our retention rate would be 47%?"

Connie Simpson: "If you go back to 1999, how many online marketing campaigns were we executing each year?"

Penny Parker: "Maybe a dozen, a monthly e-mail campaign at most."

Connie Simpson: "And today, how many online marketing campaigns are we executing on an annual basis?"

Penny Parker: "Oh, heavens, thousands, maybe tens of thousands, coupled with hundreds of thousands in search."

Connie Simpson: "And yet, our annual retention rate is the same as it was ten years ago. So what is the point of all of these campaigns? Do any of them generate incremental sales, or is all of the work done by your team just cannibalizing work done by other members of your team?"

Bill Bledsoe: "This is a multi-channel world, Connie. Multi-channel customers are the best customers. We have to be in all of these channels, or we lose market share among the most valuable customers."

Ashley Zimmerman: "I need the creative flexibility to run the products I want to run, and I need the creative flexibility to present the products the way I want to present them. I'm not going to have some marketing bean counter telling me what I should or should not do. Do I tell the marketing bean counter how to execute a test, do I dictate the control size, do I force the marketer to use a certain type of statistical test, do I judge the analytics wizard based on one-tail or two-tail tests? Certainly not! They are the experts, they get to do what they do best. I, too, should be able to do what I do best. I am the expert when it comes to presenting products."

Connie Simpson: "And yet, Penny's data suggest that you are not the expert, her data suggests that your customer has preferences that you can follow in order to improve business."

Ashley Zimmerman: "Connie, do I tell you how to invest money? You lost $200,000 last month by hedging against the Euro, what does that have to do with selling shoes? Maybe we should have Penny run optimization tests against your hedging strategies. And Bill, do we honestly think that your delivery tracking system is working best? Maybe we should have Penny measure how many days it actually takes to get a product to a customer, and then optimize against UPS or USPS or FedEx. How does that sound? I mean, seriously, everybody loves measurement and optimization until you are the person being measured and optimized. Then it isn't so much fun. Lauren, I need the creativity to do what I think is best."

Penny Parker: "And Lauren, I think it is sub-optimal to present merchandise the way Ashley wants to present merchandise, I think it is hurting our business."

Bill Bledsoe: "Kevin, you are sitting there smiling, why?"

Kevin: "Lauren, what do you think we should do?"

Lauren Fetzer: "I think we need to give Ashley the room to do her job as she sees fit, and when she fails, she's fired. Penny is free to measure Ashley's techniques, but by the same token, Penny is accountable for her own campaigns, and as best I can tell, her campaigns aren't working, because our annual retention rate isn't moving. So in time, Penny and Ashley will succeed, or they will fail."

Bart Cox: "That's how we do things in retail. We don't need to be hit over the head with a bunch of complex metrics. Either things work, and you keep your job, or they don't work, and you're fired."

Kevin: "That's why I was smiling. Lauren is playing the role of ..."

Penny Parker: "The HiPPO. She always plays the role of the HiPPO."

Kevin: "Yes, and that role, despite being despised in the social media and analytics world, is needed for dispute resolution."

Bart Cox: "We don't have disputes in retail. We do what I say we should do. My team marches in formation to my commands."

Penny Parker: "But that's a problem with this company. We're not a data-driven company. We have the metrics to prove points, and then Lauren's opinion trumps others."

Ashley Zimmerman: "We need to be a merchant-driven company, Penny. Without merchandise, we are out of business. Without marketing, we aren't profitable."

Lauren Fetzer: "We'll, we've beaten this horse long enough, it is time to move on to something productive."

Penny Parker (whispering to Kevin): "HiPPO".

One of our loyal readers, named "Anonymous", forwarded this word cloud of the MineThatData Blog.

Word clouds, of course, convey the phrases that are prominent on the blog, coupled with words that are used in association with each other.

Notice that Lauren Fetzer gets a disproportionate amount of space!

To create your own Word Cloud for any website you commonly read, visit Wordle.

Online marketers like to segment by first time visitors, prior visitors, prior buyers, that kind of thing.

Catalog marketers like to segment by recency / frequency / monetary values.

Give this one a try. Take all of your first time buyers who purchased two items. Split those customers into those who bought from just one merchandise division, and those who bought from multiple merchandise divisions. Then measure the long-term value of each segment of customers.

It turns out that, in most cases, customers buying a broad range of products are more valuable than customers who have a preference for just one merchandise division.

So give that a try ... and if you are a web analytics expert, take a look at conversion rates by this level of segmentation!

Dear Catalog CEOs:

We're now being told that a postage increase is looming ... maybe 5%, maybe 10%, who knows?

Embrace the increase.

Honestly, I don't understand why so many of us view this as yet another crippling blow to the catalog marketing model. We rally around each other, we mobilize (to some extent, though most of us won't join the ACMA and actually put a few thousand dollars behind the effort, so do we truly care about this issue?), we get ready to fight.

Most of all, though, we complain, we fret, and we worry.

Here's the thing, folks. There is this thing called "the internet". Odds are that the internet is most of the way through the process of crippling the existing system of delivering information through the mail.

The internet gives us an out. We have other ways to sell merchandise. Instead of fighting tooth and nail to preserve the old, why not truly scare those who appear to be persecuting us with a five percent increase?

Tell the postal service and the vendor community that supports the postal service that unless they make this form of delivery viable, we will move our businesses online, and as a consequence, they will get nothing from us.

Nothing.

Instead of clinging to the past, why not demonstrate to the postal service and the vendor community that supports the postal service that we have a vision for the future? Isn't that our job, as business leaders, to have a vision? Negotiating about whether an increase should be 3% or 5% or 7% is not visionary, folks.

Show them that your business, while smaller, will still be viable without print. Show them that you will use other delivery channels, radio, cable television, e-mail, YouTube, social media, mobile, paid search, re-targeting, affiliates, word-of-mouth. Acknowledge that none of those channels, when added together, are as effective as print is today, but demonstrate that you will learn and you will grow, and the insight you will gain will allow you to eliminate your print program altogether.

And when you eliminate your print program altogether, the USPS and the vendor community that supports the USPS gets nothing. Their cost increases won't increase revenue, their cost increases will result in a situation that hastens their demise.

Challenge the USPS and the vendor community that supports the USPS to make their delivery channel relevant and competitive in an era where digital delivery of information is far more efficient than print and is virtually free. Don't operate out of a position of fear. Be confident.

Of course, I am here to help you through this transition. Contact me now for assistance.

Thanks,
Kevin

Today, I am meeting Lauren Fetzer, CEO of Fetzer's Footwear, at Eagle Heights, a barren prairie on the southern end of Madrona Island. As usual, Ms. Fetzer is listening to her iPod Touch.

Kevin: "What are you listening to?"

Lauren: "Walking on Broken Glass by the Eurythmics" That's how I feel when I'm dealing with my creative team, you have to tread lightly with them."

Kevin: "Why are we meeting here?"

Lauren: "Look around, there aren't any trees here. There were trees at one time, but they were harvested a hundred years ago. Now, they have a hard time getting started again. In November and December, the windstorms down here are brutal. So the landscape changed, it evolved."

Kevin: "What would it take for a tree to grow here again?"

Lauren: "Well, it would have to really want it."

Kevin: "What does that mean?"

Lauren: "A friend of mine has a saying. She says 'you have to want it' when it comes to success. I couldn't agree more with any other statement than that one. The tree would have everything working against it. The soil is sandy now, so if it did grow to a reasonable height, the winds in November and December would uproot it. The tree has to want it!"

Kevin: "It wouldn't be a best practice to start growing here, would it?"

Lauren: "Absolutely not. But if the tree made it, and the odds are against the tree, it would pave the way for a forest. Nobody would ever recommend that the tree take root here, would they? No, the experts would find the perfect environment, they would take soil samples and they would measure everything so that they could find the best place to plant the tree."

Kevin: "And they would be successful, wouldn't they?"

Lauren: "Define success? The tree would grow, but it would be one of, what, a million trees? Is that success? Who wants to be part of a forest?"

Kevin: "What does this have to do with Fetzer's Footwear?"

Lauren: "Have you ever taken part in strategic planning, Kevin? I assume you have."

Kevin: "Sure."

Lauren: "Strategic planning is not a spreadsheet exercise, Kevin."

Kevin: "Meaning what?"

Lauren: "Penny Parker, you've met her, she is our VP of Marketing, she has all of these fancy spreadsheets and metrics. Her life is a series of campaigns. She'll line up two hundred different campaigns for next year. She'll figure out how to optimize every single campaign, to squeeze an additional eight percent out of every single campaign. Then she sits back with this big smile. She adds up eight percent on top of eight percent on top of eight percent and thinks she has the answer. All we have to do to grow the business is be perfect across two hundred consecutive campaigns, and if we're perfect across each and every one of two hundred campaigns, our business will grow by something like thirteen percent. She has the metrics to prove it, she opens up this big spreadsheet and puts it up on a plasma monitor and everybody gazes as all of the metrics, and she is manipulating one cell and twelve other cells all improve. Wow. She's really optimizing the business, isn't she?'

Kevin: "Why the attitude?

Lauren: "Well, for one thing, her strategies never work. Who can be perfect two hundred times out of two hundred? If you do your best, you might improve the performance of one hundred and fifteen campaigns, and you'll inadvertently fail eighty-five times."

Kevin: "That's still improvement, right?"

Lauren: "It's fools gold. Her answer is to simply execute more campaigns, if you have more campaigns, you'll have more sales, and if you have more sales, your business will grow. Paid Search, Email Marketing, Affiliate Marketing, Organic Search, Print Ads, Banner Ads, Re-Targeting, Radio. Now we have thousands of campaigns, and our marketing team and analytics team believe they can optimize each and every one of thousands of campaigns, squeezing an additional four percent and six percent out of each one. It's funny, Kevin, nobody in marketing ever says that they expect campaigns to perform less well. Every single metric in every single spreadsheet points north, showing improvement. Fools gold."

Kevin: "Maybe you hate metrics and spreadsheets? There's lots of people who feel that if you test and optimize your business will experience unfettered growth. They are frustrated by people like you."

Lauren: "We had this consultant in a few weeks ago, maybe you know him, his name is Chip Cayman. He had this document that said if we grew to a hundred stores, our sales would essentially quadruple, his spreadsheets showed that multichannel customers are worth nine times as much as single channel customers, he showed how the website would benefit from a national store presence. Nice presentation, classic stock photos that he paid $6 each to obtain, great math, an opportunity to rack-up ten or twenty million dollars of debt that a spreadsheet suggests will be paid off with profits from the stores in ten years."

Kevin: "So what is the problem?"

Lauren: "He didn't want it."

Kevin: "He didn't?"

Lauren: "Not even close. He wanted somebody to love his spreadsheet, his math. He didn't give a rat's behind about the relationship between a customer and an employee ... it is a humbling experience to serve a customer who wants to buy shoes in a store. No spreadsheet solves for a human being wanting to help another human being."

Kevin: "Why are you telling me this?"

Lauren: "Tell me I am right."

Kevin: "So what if you are right? You are the CEO, it doesn't matter."

Lauren: "It means everything. Penny hates it when I ride her on this topic, she calls me a 'HiPPO'"

Kevin: "A hippo?"

Lauren: "Highest Paid Person's Opinion. She has all of this science, and she believes that if I just listened to her science, everything would be fine. Then I tell her that she 'has to want it'. I tell her to have a passion for our business, not a passion for spreadsheets and optimization. And she just stares at me, she says the spreadsheet tells the whole story. But listen to me, Kevin. She's been optimizing year after year after year, and somehow, our annual retention rate, the percentage of customers who bought last year and will buy again this year, that never budges, it always stays the same. Well don't you think that if she was right, if we listened to her and her glorious metrics, that the annual retention rate would improve? See, that's the problem. If all of that math and optimization were right, business would improve. It doesn't improve. She doesn't want it, she wants science."

Kevin: "So what does wanting it look like?"

Lauren: "Wanting it means being willing to take a risk. There are no risks in spreadsheets, only numbers. Wanting it means innovating. Wanting it means moving beyond campaigns. Campaigns are nothing more than glorified begging, you are trying to manipulate the customer into buying something."

Kevin: "Isn't that what a marketer does? Don't marketers manage campaigns?"

Lauren: "Marketers tell stories. See, that's what matters. A marketer that doesn't try to tell a story is a marketer that doesn't want it. Here's the neat thing. If you tell a story, and if you are consistent with that story, and then the sales naturally follow."

Kevin: "Or they don't."

Lauren: "And if they don't follow, it is self-evident that the story didn't resonate with the customer. Then you try another story. The marketer must have an instinct for getting from Point A to Point B. Look, we're all getting killed by Zappos. They tell a story. Their story is flimsy, of course, but it is a story. They ship you product fast, they have every sku imaginable, they have perceived free shipping, and their employees will chat with you via social media. That's a story. There's no optimization and spreadsheet manipulation and campaign management there, heck, where was the best practice for that business model before they invented it? Coming up with that story, a marketing story supported by merchandise, that's hard work, Kevin, it isn't easy. That's what I call 'wanting it'. We have to want to beat Zappos so bad that we come up with a better story. When people don't have a better story, they revert to campaigns and optimization and spreadsheet manipulation."

Kevin: "So the tree trying to grow on this prairie needs a different story?"

Lauren: "Exactly!"

Kevin: "Seems like it would be easier to simply transplant this tree in the forest on the other side of the island."

Lauren: "Yes, it's easier to manage by spreadsheets and metrics and campaigns and optimization. That's not what I am looking for. We need to break through, we need to tell a story that is more compelling than the story that Zappos tells. We need to want it more than them."

Kevin: "Yup, you're just another dumb HiPPO."

This summer, we're going to talk a bit about segmenting customers.

See, it turns out that this is one of the most important times in history to actually slice and dice customers. Our customers are going in a myriad of different directions. Back in 1995, you could do a simple RFM analysis, and you'd be in good shape, because you had a one-dimensional business.

In 2010, your customer is a few years into the early stages of a great diaspora ... a once homogeneous audience is now dispersing itself across a nearly infinite number of micro-channels. What is important, here, is that while customers adopt an increasing number of micro-channels, overall response does not improve --- you know this because you look at annual retention rates, and your annual retention rates have not improved in a decade. When the number of micro-channels increase and overall response fails to increase, you have tremendous "optimization" opportunities!

So, we're going to explore segmentation this summer. Get ready!

Sometimes, when searching, you really want to get a hold of the URL of one of the documents in the result set. Generally, the URL of a search on Google is displayed below the title and snippet of the result. However, it is not uncommon for the URL to be shortened in some way if it is too long, so that you can at least see the host name and the name of the file.

 

When this happens, however, you can't copy and paste the URL for use elsewhere. The link backing the title, which you click on in order to get to the document isn't actually the URL of the document, but a proxy that Google users to track which users click on what, so copying that URL won't get you the URL of the document either. The URL is in there, but encoded.

One way around this problem is to click on the 'similar' link. This will take you to a Google results page for which the query (found in the search box) is composed in part of the link you are after. You can then copy it from there.

 
Ideally, from a user perspective, Google would back the URL displayed in the results with the actualy URL. 

Dear Catalog CEOs:

Twice in thirteen days members of your elite club reached out to me to communicate interesting findings. In one case, one of you stopped mailing catalogs altogether and retained 90% of your sales one year later (pure catalog brand, no retail channel). In another case, one of your staffers executed a catalog holdout group for more than nine months, and retained 90% of your sales within the holdout group (pure catalog brand, no retail channel).

It's not a bad thing when you stop mailing catalogs and sales are generated anyway. In fact, it's the very best thing that can happen. YOUR CUSTOMERS ARE BRAND LOYAL!

Do you love sending catalogs, or do you love printing money? So often, the answer is the former, not the latter.

Any business that maintains sales when advertising stops is a blessed business. Run a mail/holdout test, and see how much brand loyalty you have!

While Nicholas Carr argues that the internet actually affects the brain making individuals dumber, Clay Shirky only argues that the internet creates useful cultural artifacts (regardless of whether or not something bad is happening to your brain). This pair of articles is definitely worth reading, but they don't appear to meet in the middle. I suppose one could read more in The Shallows: What the Internet Is Doing to Our Brains, or Cognitive Surplus: Creativity and Generosity in a Connected Age, two books which happen to be mentioned at the end of the respective articles.

From Multichannel Merchant ... here's the quote about the closing of Abbey Press:

• "All businesses have a lifecycle, and the monks felt that the catalog business was past its peak as a dominant business model".

And to top it off:

• Wilson says Snail’s Pace will reflect changing marketing paradigms: “You’ll see less money spent on advertising, catalogs, direct mail and the like, and more effort put into reaching exactly the right customer prospects through social media channels, the Internet, and through media that they are currently consuming.”

Now half of you will say that "... print isn't dead, my vendor/co-op just showed me matchback results and 81% of my sales are attributed back to the catalog." And you may be right, assuming that your matchback results tie out to your mail/holdout test results. Have you done the analytics work to prove that your matchback results are accurate? Has your vendor recommended mail/holdout testing to validate matchback results?

If your matchback vendor hasn't recommended executing mail/holdout testing to validate matchback results, be afraid ... be very afraid.

I'm not saying you shouldn't be mailing catalogs ... I'm saying you should make profitable decisions. Look at what is happening all around you, and pay close attention to how customer behavior is shifting.

Want to make a good use of the next ten minutes? Give this transcript a read ... J. Crew pumped-out 18% pre-tax profit last quarter. In case you didn't know, that's not an easy accomplishment. Ten percent pre-tax profit is often considered a gold standard in apparel retailing ... in my sixteen years at Lands' End, Eddie Bauer, and Nordstrom, the companies I worked for only exceeded 10% four times (I think three of the times were at Nordstrom). Just read the merchandising passion they exude in a boring investor conference call.

And then there is Rue La La --- debuting apps: Don't tell me that "... that doesn't apply to us, we're not some low price auction brand, our customers are different, they love sitting in front of the fireplace thumbing through a catalog that has been sitting on the coffee table for twelve weeks" ... that's an excuse to not even try, folks. It is pretty obvious now that the future isn't e-commerce. If I am running an e-commerce division, I want my customer to hold my brand in her hand at all times ... anything short of that and we're just clinging to the past.

Fascinating to see Google' come out with a Bing-like image background.

Note that some work is needed on the fit and finish to avoid readability issues like this:

I enter Lauren Fetzer's office. As always, she's listening to her iPod Touch.

Kevin: "What are you listening to?"

Lauren: "Bonnie Raitt: 'I Can't Make You Love Me', it reminds me of things that happened in college."

Kevin: "I'm sorry."

Lauren: "Moving right along, today you're going to get to meet with my Executive team. We'll have you sit in on some of our meetings. Feel free to offer your opinions, I won't censor them. And don't let anything Bart Cox says offend you. Every good Executive team should have one crusty individual with a contrarian point of view. All viewpoints are welcome at Fetzer's Footwear. Let's go."

We enter the Executive Boardroom.

Lauren Fetzer: "Good morning everybody. I want to introduce Kevin Hillstrom to you. Kevin will join us from time to time to review our current strategies, analytics, and help us achieve our goal to craft a vision for the future"

Bart Cox (VP, Stores): "Who is this guy, again, no offense?"

Lauren Fetzer: "Kevin is a direct marketing industry analyst, and Bart, Kevin previously worked at Eddie Bauer and Nordstrom, so he does know something about stores."

Bart Cox: "Sure he does."

Lauren Fetzer: "Let's start our meeting. Penny, please share business results from yesterday."

Penny Parker (VP, Marketing): "Comp store sales were up 2.8% yesterday, led by Bellevue with a +8.3%. Alderwood was at -6.5%. The website was up 14.5% from last year, and was up 9.5% vs. budget. The Mountain-ariffic e-mail blast was responsible for the entire increase yesterday, that was an outstanding campaign."

Lauren Fetzer: "Bart, what happened at Alderwood yesterday? That's three consecutive weeks of negatives at Alderwood?"

Bart Cox: "We've talked about this for two consecutive weeks, the mannequins are too athletic, and as a result, we're not drawing customers into the store."

Ashley Zimmerman (VP, Merchandising): "Bart, we have the same mannequins in Bellevue and Bellevue is running positive comps."

Bart Cox: "The Bellevue customer is different, she is more likely to hike in the Cascades. The Alderwood customer is functional, she's going to wear her shoes to Red Lobster. You already know this, just look at the taper reports that show what sells by store, the best selling styles are totally different."

Bill Bledsoe (VP, Logistics): "Penny, do you see these customer differences online? In other words, do you have online taper reports that show what sells by geography?"

Penny Parker: "Now that's a great idea, we'll go back and take a look at that!"

Connie Simpson (VP, Finance): "Year-to-date, retail is trending at +3.3%, online is +11.4% to last year at +6.8% to budget. I'd say that we're having a very nice year. Your teams deserve kudos for their hard work."

Lauren Fetzer: "Let's talk a bit about our vision. Penny, what did you learn about where the experts think online retailing is headed?"

Penny Parker: "I purchased a Woodside Research report that predicts 88% of businesses will generate up to 22% of their revenue in 2015 from a combination of mobile and social."

Bart Cox: "You know, folks have been saying all of this virtual stuff will wipe out retail for at least fifteen years now. And it never does. Retail keeps plugging along. Last time I checked, retail comprised 90% of all purchases, just like it did fifteen years ago. So, yes, keep thinking visionary thoughts, folks, but remember that from a net sales standpoint, almost nothing has changed in the past fifteen years. Back then it was telemarketing and direct mail, then it was the internet, now it is mobile and social."

Ashley Zimmerman: "That doesn't change the fact that we have to change with the times, Bart. You probably didn't envision a customer walking into your Downtown Seattle store with a Fetzer's Footwear iPhone app back in 1995, did you? Today she walks into the store, and if the item she saw online isn't available in the store, she punches up her app and gets the item shipped to her home. So you have to adapt and change."

Bart Cox: "And I want credit for that sale, too, because I generated that sale. Bill, when are we going to make the reporting changes necessary to show that I drove that sale?"

Bill Bledsoe: "Who cares who drove the sale, just be thankful that somebody bought something! In that case, the website, the store, and the app all contributed to the sale. That's the future. Trying to parse out that combination and assign weights to each channel is nothing short of fool's gold."

Bart Cox: "Or a lack of analytical imagination."

Lauren Fetzer: "So the future involves more channels and more analytical complexity. That's what we need to focus on. Kevin, where do you see the customer in five years?"

Kevin: "I think she'll have moved away from static websites. I believe that the traditional e-commerce website is already a 'dead man walking' if you will. So much of what a customer does today has nothing to do with the traditional e-commerce website. She's on Facebook, Twitter, Foresquare. He's on a mobile app that basically shifts eyeballs away from a traditional website. All of the things we deal with in e-commerce, our entire knowledge base, is dying. We need to learn to be really good at being there for our customer. When she needs us, we're there. When she doesn't need us, we quietly disappear. We cannot assume that she will visit us like we are some sort of destination that she desperately needs."

Bart Cox: "That's just hokey theory."

Ashley Zimmerman: "Sure it is, but we need to think about it, because you can see it happening already. We worked so hard to get user generated comments on our website, only to realize that we get a hundred comments on Facebook and Twitter for every user generated comment on our own website. User generated content on a website is so 2008. My customers love my product, and they talk about it all of the time in their social circles, they don't come back to our site to chat with us, we have to go out there to chat with them."

Bill Bledsoe: "I don't want brands following me out into my 'solar system', if you will, to chat with me, that's just plain creepy."

Connie Simpson: "So that speaks to our future, doesn't it? How do we follow the customer out into the 'cloud' without announcing that we are right there next to her, screaming at her every five minutes to buy something?"

Bill Bledsoe: "That's why I like our our series of apps for all mobile devices. We're essentially following the customer in an invisible manner. When she needs us, she just punches us up!"

Penny Parker: "That's where my job becomes challenging. Somehow, I have to encourage the customer to carry our apps with her without sounding like a snake oil salesman."

Lauren Fetzer: "Good discussion, folks. I want for us to honor our meeting requirements. We have ten minutes left before this meeting ends, and as you all know, we end all meetings at five minutes to the top of the hour so that nobody is late for the next meeting. Thanks to Kevin for being here, he's going to join us a lot in the future as we flesh out where we think we want to take Fetzer's Footwear."

Bart Cox: "I think we want to open a bunch of stores!"

Tomorrow (Thursday) at 3pm EDT I'll be on DM Radio for the broadcast "Embedded Analytics and Business Rules: The Holy Grail?".  I'm not sure what the other guests are going to talk about, but my comments will resemble the talk I gave at Predictive Analytics World in February 2010 in the talk Rules Rule: Inductive Business-Rule Discovery in Text Mining. In this help-desk case study, we used decision trees to cherry pick interesting rules, converted them to SQL, and deployed them in a rule system that was applied transactionally, online. I emphasized the text mining portion at PAW, but the methodology was independent of that. In 2002-2003, researchers and I at the IRS applied same kind of approach to rule discovery in selecting returns for audit: use trees to find interesting rules.

The reason we liked the approach was that it was a fast way to overcome two problems. First, decision trees only find the best solution to a problem (according to its measure of "good"). To obtain a richer set of terminal nodes, one can build ensembles of trees, but then one loses the interpretation. On the other hand, one can build association rules, but then you are left with perhaps thousands to tens of thousands of rules that have to be pruned back to get the gist of the key ideas. Many of the rules will be redundant (some completely identical in which records are "hit" by the rule), and it's easy to become lost in the sheer number of rules.

For the Fortune 500 company, we used CART with the battery option to generate a sequence of trees (we iterated on "priors" and misclassification costs, and I think some more options as well to generate variety), and took only those terminal nodes that had sufficiently high classification accuracy. I think we could have used their hotspot analysis for this too, but I wasn't sufficiently well-versed in it at that time.

If you can't join in on the radio broadcast, you can always download the mp3 later.

Just like last summer, we'll go on a reduced schedule for the summer. Traditionally, blog readership peaks in February and March, then begins a slow decline that culminates in minimal traffic in late July.

From June 7 - September 6, we'll go with a Monday / Tuesday / Thursday schedule, with additional posts on Wednesday as opportunities present themselves.

If you are craving additional updates, come on over to Twitter and follow me there.

Dear Catalog CEOs:

Some of you are already using a prospect catalog to improve overall performance.

Here's how the concept works. Let's assume you currently have a 128 page catalog. When you mail the 128 page catalog to your prospect audience, you observe the following metrics:


128 Pages

Circulation 500,000 Demand $900,000 $/Bk $1.80 Profit Factor 37.0% Book Cost $0.64 Profit $13,000 AOV $100.00 Response 9,000 Profit/Ord $1.44
Ok, those aren't bad metrics.

Now, what might happen if we used a 48 page prospect catalog? Let's assume that the old standby rule of "pages are added at half-productivity" holds here. In that case, we can use the "square root rule" to estimate demand for a 48 page catalog:

• Demand For A 48 Page Catalog = (48 / 124) ^ 0.50 = 0.612.

In other words, we'll obtain 61.2% of the demand.

Next, a prospect catalog is merchandised with only the best products. When merchandised in this manner, productivity usually increases by 20%, plus/minus. Let's assume a 15% increase in productivity.

• Demand For A 48 Page Catalog = 0.612 * 1.15 = 0.704.

Finally, we'll make two additional assumptions. First, we'll assume that the average order value is a bit less, because there are fewer pages ... we'll assume a $95 average order value. Second, we'll assume that each page in a 48 page catalog is more costly to print and produce ... we'll assume that this catalog is 15% more costly to produce.

With these assumptions, what might a profit and loss statement look like for a 48 page catalog, vs. a 128 page catalog?


128 Pages 48 Pages


Circulation 500,000 1,340,000 Demand $900,000 $1,037,584 $/Bk $1.80 $0.77 Profit Factor 37.0% 37.0% Book Cost $0.64 $0.28 Profit $13,000 $14,066 AOV $100.00 $95.00 Responses 9,000 10,922 Profit/Ord $1.44 $1.29
There are a lot of positives here:

• You expose your brand to 1,340,000 households, not a measly 500,000 households. Your co-op will love you for this!
• Demand increases by more than ten percent. Your CFO will love you for this!
• Profit increases in this scenario. Your CFO will love you for this!
• Responses increase by more than 20%. Your Circulation Director will love you for this!

This is why you create a prospect catalog. All of the metrics that matter improve, allowing you to greatly increase circulation depth, to increase demand, to increase orders, and to increase profit.

See, I'm not all about cutting catalog marketing, am I?

I love the Zoom books by Istvan Banyai, and I really like the SIFF trailer which animates the Zoom concept over classic cinema moments:

While to most of us, the first name that comes to mind when we hear about the oil spill is BP, as this article in the New York Times indicates, there are a number of commercial entities involved, including BP, Transocean, Schlumberger and Haliburton.

Bloggers are spending most of their attention on BP.

The stock market appears to have divided their attention more evenly (among those entities in the public domain), with Transocean (RIG) taking the deepest dive.

 

The PAKDD-10 Data Mining Competition results are in, and ensembles occupied the top 4 positions, and I think the top 5. The winner used Stochastic Gradient Boosting and Random Forests in Statistica, second place a combination of logistic regression and Stochastic Gradient Boosting (and Salford Systems CART for some feature extraction). Interestingly to me, the 5th place finisher used WEKA, an open source software tool.

The problem was credit risk with biased data for building the models, a good way to do the competition because this is the problem we usually face anyway: data was collected based on historic interactions with the company, biased by the approaches the company has used in the past rather than having a pure random sample to build models. Model performance was judged based on  Area under the Curve (AUC), with the KS distance as the tie breaker (it's not everyday I hear folks pull out the KS distance!).

One submission in particular commented on the difference between how algorithms build models and the metric used to evaluate them. CART uses the Gini Index, Logistic regression the log-odds, Neural Networks minimize mean squared error (usually), none of which directly maximize AUC. But this topic is worthy of another post.

I’m continually impressed by the breadth of content presented at ICWSM. This conference is truly achieving a unique position for social media and social network researchers: bringing together social science and computer science. I thought it would be interesting to capture the conference visually, so below is a selection of the visual representations of data and results from some of the papers from the conference. Each is linked to the online paper page, so click through to learn more!