The Optimization Arc – From Black Box to the Cloud

Joshua Trees

WHEN THE GOOD FOLKS at DemandTec asked me to commence writing a series of short commentary pieces on this blog, I accepted the assignment in large measure because the company’s story is a reflection of the story of merchandising analytics in all its facets.

Like DemandTec, an IBM Company, my history as an analyst in what used to be called the “price optimization” sector dates back more than a decade. In 2002 I was asked to try to make this very powerful new retail science more accessible by explaining its benefits and justification in terms other than technical. Price optimization was a new idea, and its target purchasers were wary of its mysterious mechanisms.

Retailers’ objections about the apparent “black box” nature of base price elasticity permeated the sales cycles of industry pioneers, DemandTec’s included. Prospects worried that using computers to model price elasticity and interaction effects to maximize margins was too manipulative. What kind of push-back would they face if shoppers found out?

It took some effort at first, but we correctly reasoned that since optimization is based on measurements of shopper response, it is inherently shopper-centric in nature. Overall, the process tends to deliver more consistent competitive value to shoppers, while retailers maintain sustainable gross margins. These ideas are familiar now, but they were new territory ten years ago.

At about the same time, other pioneers began applying the principles of optimization to other complex merchandising decision processes, notably to the depth and timing of markdowns, and the terms of in-store promotions. Other folks were advancing assortment and space planning tools from the category management side of the house. Pretty soon, it dawned on the smarter people that that the interconnectivity and interaction effects they observed within each of these areas of discipline also exist across these areas of discipline; and not just within the retail organization, but between it and its trading partners.

A simple example might arise when a lower everyday price for a popular item revs up its turnover rate. The existing number of facings may become insufficient, creating intermittent out-of-stocks. The lost sales may tend to distort apparent demand and delay re-orders, and the problem perpetuates. Fold in other concurrent events within the category, such as new item cut-ins and shelf capacity constraints and the problem grows very knotty indeed.

Fly by Wire

When I was first learning about all this, someone I respect explained to me why the mathematical model behind pricing optimization is related to the intricate “fly-by-wire” flight control systems that keep stealth aircraft from dropping out of the sky. Both critical objectives – keeping thousands of interrelated SKUs properly tuned, and keeping multiple interrelated flight surfaces properly tuned – share several traits:

  • The model is big
  • The model must be dynamic and continuous
  • The model must be highly reliable under duress
  • The model must be continuously updated at a time cycle that is rapid enough to support critical decision-making
  • The model must be appropriately accessible to decision makers

In one respect, those flight control systems may be simpler than retail demand models – there’s only one cockpit in an aircraft. A retail organization, by comparison, may have dozens or hundreds of individual decision makers and planners and trading partners interacting with the merchandising model through various dashboards. Each needs appropriate analytics and decision support according to his or her role.

To the Cloud

As DemandTec developed and acquired its portfolio of software offerings over the past ten years, it placed evident emphasis on connecting users with the data and with each other in practical and beneficial ways. It was an early advocate of the software as a service (SaaS) application business model, which placed the heavy application power in outside computer servers, relieving clients from the burden of maintaining these systems in-house.

Lately the tech industry tends to refer to service-based computing as “the cloud.” In fact DemandTec’s current positioning, “The Collaborative Analytics Cloud,” reflects that. The explosive growth of major social networks has reinforced this concept, as have some of the largest IT companies. IBM, which acquired DemandTec last February, uses the tagline, “Smarter Commerce on Cloud” to describe its core strategic approach.

The company’s DemandTec Connect™ social layer is a recent development in this regard. The platform leverages social-media-like interaction with embedded analytical applications to help shape collaboration across the merchandising ecosystem. Like any social media network, the platform is cloud-based. Its ability to provide role-appropriate access to a variety of optimization analytics is pure DemandTec.

© Copyright 2012 James Tenser
This article was commissioned by DemandTec Inc. which is granted the right of republication. All other rights reserved.

Call It Mobile Yellow

Scan these tags!

I SPIED THE WHOLE grocery universe in a single bunch of fruit. This transcendental experience occurred just last week in West Des Moines, IA. (Quite rightly.)

It all happened in a bright, spacious HyVee supermarket in an upscale, verdant neighborhood that a decade ago was pretty much a cornfield. I entered the store thinking, “They built this, so I have come.”

But my Field of Dreams reverie dissolved when I stepped up to the produce department. My first view was an abundant display of bananas – the largest selling produce item in most supermarkets.

I pulled out my smartphone and snapped this picture, with the classic lyric by Donovan resonating in my brain like a prophetic soundtrack:

Electrical banana is gonna be a sudden craze
Electrical banana is bound to be the very next phase

You see, what really blew my mind were the two little stickers on this hand of bananas. On the left, a stacked UPC bar code and on the right a QR (quick response) 2D code. Each of these artifacts tells a little story about the impact of digital technology on our business. (Quite rightly)

The UPC bar code is of format known as RSS_14 designed to convey more information than old-style 9-digit codes. When I scan it with a gadget on my smart phone it correctly identifies the product as Dole bananas. Presumably the POS scanners in this HyVee supermarket are just as smart. The advantage: it saves the checker a few keystrokes and accurately enters the PLU (price look up) code for this variety of produce which is sold by weight.

The QR code on the right is used to direct shoppers to a web site called yonanas.com . When I scanned it with the same smart phone app, it led me to a page pitching a simple kitchen appliance that lets consumers make a soft-serve dessert from frozen ripe bananas and other fruit.

The Dole brand is featured prominently on the target Web page. Presumably there’s a deal behind it all. The Yonana appliance is marketed by Healthy Foods LLC, itself a division of Winston Products LLC in Cleveland, Ohio.

Peering at my smartphone screen under the fluorescent lights in that immaculate HyVee store in one of the greenest places in America, I found myself thinking, “It really has come to this: agricultural and digital have converged in America’s heartland.”

Then the digital banana concept set off a minor cascade of linguistic play: Bananas come in hands… Digit is Latin for finger… Smart phones are made with gorilla glass… (Quite rightly)

So in tribute to Donovan, the iconic Sunshine Superman inducted this year into the Rock and Roll Hall of Fame, I’d like to offer this (somewhat less lilting) variation on his theme:

Digital banana is gonna be a sudden craze
Digital banana is bound to be the very next phase
I call it mobile yellow (Quite rightly)

© Copyright 2012 James Tenser

A Web of Truths

WATCH OUT, Shopper Marketers! You may find yourselves entangled in a web of truths of your own making.

It all began innocently enough; in 2005 when brand marketing behemoth Procter & Gamble advanced a provocative set of ideas around what it called the first and second moments of truth. Thanks to some savvy and persistent promotion, the terminology caught on fast:

  • FMOT, the first moment, refers to the brief period when a shopper selects a desired product in the store.
  • SMOT, the second moment, refers to the at-home consumption experience associated with that product.

Within the then-nascent Shopper Marketing community, this framework was a minor revelation. For brand marketers, FMOT gave credence to the argument that real marketing persuasion needed to be extended from measured media into the shopping environment. The store, it was discovered, shelters a separate marketing reality, where pre-purchase leanings are transformed into final choices.

Shopper Marketing defined a path to purchase that commences with media-induced product awareness and proceeds to interest, formation of intent, and ends with product selection at the shelf, FMOT. Once home, SMOT, or the actual product experience, takes place influencing subsequent decisions.

FMOT/SMOT was a pretty handy framework at first. But the concurrent rise of digital out of home and mobile media conspired to make things a lot more complicated, fast. The path to purchase, it turns out, is littered with hundreds of moments – text messages, in-store video ads, Web search, service encounters, Facebook apps, twitter feeds, QR codes and downloadable coupons, to name a few.

Stuck in the Moments

A few weeks ago the gleefully disruptive folks at Google seized the opportunity to coin a new Moment of Truth and promote it hard. They call it Zero Moment of Truth or ZMOT. Its premise is that interactions with search, Web, social and mobile price and product research media create a third type of online decision-making moment. The concept is a bit self-serving coming from the world’s largest seller of online advertising, but it has attracted much commentary and attention.

Almost immediately, new Moments starting appearing like so many pop-up windows on an e-commerce Web site.

In his post, “What is missing from moments of truth marketing”, blogger Joel Rubinson argues for the existence of “minus one” moments of truth that include such influences as word of mouth, in-store product visibility, and various types of advertising. Most interestingly, he proposes that these -1MOTs may occur in any sequence relative to FMOT and SMOT.

Joel’s point about the non-linear nature of the Moments of Truth is worthy of frequent repetition. Product experience is certainly a web of moments, not a fixed linear sequence. Call it WOT (Web of Truths)?

On the very same day and from an independent thought process, blogger David Berkowitz proposed adding “The Infinite Moment of Truth” to the model, which reflects his excellent observation that consumers may well describe their product and service experiences to others, relaying and amplifying the message beyond the scope and control of the marketer.

Bon MOTs

I applaud David for extending this Shopper Marketing discussion from the path-to-purchase toward the path-to-loyalty. A good thing, really, since the linkages are powerful and real. It made me think about Fred Reicheld’s 2006 book, The Ultimate Question, which proposed that genuine loyalty was best judged by an individual’s likelihood to recommend a product or service to others. Social media can super-charge this potential.

Both bloggers are smart, experienced people I know for some years and their ideas are intelligent and worthy of respect. But I must confess to an impish reaction that led me to ponder: Just how many bon MOTs can one industry handle? ZMOT; FMOT; SMOT; Rubinson’s -1MOT; Berkowitz’s IMOT…

At risk of attracting ridicule, my imp compels me to toss another acronym into the mix: XMOT, the eXtended Moment of Truth. It’s my way of stretching the Web of Truths a bit wider – not quite to infinity, but toward its potential to help us understand the multifaceted tangle of influences each person receives, reflects and responds to in their roles as shoppers, consumers, and friends.

© Copyright 2011 James Tenser

Pay Cycles: When Month Outlasts Money

I WAS STRUCK to read comments a couple of months ago by Walmart CEO Mike Duke who stated that the chain’s shoppers seemed lately to be running out of money in the waning days of the month. He cited the shrinking size of market baskets as evidence. Tough times leading to tough choices.

Separate recent reports about the worrisome state of our consumer economy observe that budget-conscious shoppers tend lately to purchase smaller package sizes near the end of their pay. This, of course, is a key contributing factor to smaller baskets. William Simon, Walmart U.S. stores chief, made reference to this “paycheck cycle” at a recent analyst meeting.

This morning a report in Bloomberg News described shoppers upping their use of credit cards for purchase of household necessities and gasoline. This is a confounding signal that looks on the surface like a rebound in consumer confidence. In fact, it seems to be concentrated at the end of the calendar month. This may be a sobering sign that many households’ flat and declining paychecks can’t keep pace with price increases.

I’ll leave the economic and social import of this behavioral trend to the true experts. But I would like to offer a few thoughts about the time-based shopper insights that allow analysts to detect and measure the trend. Looking at detailed market basket trends day by day, it seems, can reveal a great deal about short-term household economics.

Not Card-Sharp? Then Be a Basket Case
This is interesting because we hear a different tune about insights from the many advocates of frequent shopper programs, a.k.a., loyalty cards. The detailed segmentation data these programs can deliver offer a wealth of target marketing opportunities for retailers and their suppliers, along with behavioral insights so detailed and profound that we don’t always know how to apply them in practice.

This is very cool stuff and it is credited with upping sales and profits at some pretty sharp retailers, like Kroger. Card-linked data allows marketers to put together a picture of a whole customer relationship over time, evaluate it, and group customers into target-able groups. Walmart and the so-called “dollar” stores, however, do not go in for those card marketing schemes. They stick deliberately to their EDLP guns instead, and resign themselves to data-poverty.

Or so it may seem. Actually, there is a great deal that may be learned just by looking at basket trends, especially at those retailers who enjoy very large footprints and shopper penetration. Card-free chains like Walmart, Publix and Dollar General can track the transaction logs by day and by local geography to extract very meaningful insights. Even if the shoppers are not individually identified, their collective behavior reveals much about pay cycle trends on a store-by-store basis.

Here is where even “data impoverished” retailers can find basis for some global and targeted merchandising tactics. Carrying sufficient smaller pack sizes in key categories every day is one obvious response Walmart says it has pursued. Sales and events may be scheduled to coincide with payday for local large employers. Managers’ specials may be timed to hit key mid-month and end-of-month dates.

Well there you have it. It’s still a share-of-wallet game, even when wallets are growing slimmer. Walmart knows, there’s much of tactical value embedded within store transaction-logs, even where there’s no loyalty data in sight. It’s not just dollar size of baskets that may influence action, it’s also item counts, categories included/avoided, package sizes and purchase influences from outside factors.

When the month runs long, wise retailers jump on their cycles.

© Copyright 2011 James Tenser