Customer data won

In the Minneapolis winter, hot soup makes an appealing lunch. At a local eatery specializing in such fare, I learned an important lesson for IT.

This restaurant serves soup in bowls and “bottomless” bowls. Bottomless gets you unlimited refills. Here’s your challenge: As an IT analyst, devise a system to keep track of who’s entitled to those free refills. How would you do it?

Would you print a card at the register with a bar code to be scanned to verify which customers should get refills? Would you ask bottomless customers to show their receipts each time they return for another helping? Or …? Design your solution before you read how the restaurateur handled the problem.

OK, ready?

Bottomless customers get a differently shaped bowl.

Every time I eat there I wonder if I’d have found this simple solution. Ask yourself, and ask your analysts and designers, too, because if they restrict their thinking to IT, they can cost you a lot of money. Sometimes, a second type of “bowl” can replace a million lines of code.

Whether you’re selling soup or silverware, hardware or handbags, you’re in retail. And although IT has tremendous importance in the retail back office, in the store itself its importance is limited, which is ironic because CRM (customer relationship management) is one of the hot applications of IT these days, and most retailers live or die on good customer relations.

Take data warehousing, one of the most important tools in what we usually think of as a CRM implementation. You can use data warehousing to collect terabytes of customer information and slice it, dice it, cross-correlate it, and do multidimensional scaling if that will teach you anything.

When you’re done, you’ll learn (among other things) which customer segments, and maybe even which individual customers do and don’t buy what products from you.

Pick up a copy of Paco Underhill’s excellent book on retail, Why We Buy, and you’ll see the importance of watching shoppers shop. If you watch shoppers in action you’ll learn something data mining can’t teach you: Why (for example) your older customers aren’t buying concealer from you. It isn’t their demographics or your branding. It’s a merchandising problem: You’ve placed the concealer

on the bottom shelf. Shoppers have to bend down for it, and it’s in a high-traffic area where other shoppers brush by them.

For older shoppers bending down is bad enough. Being “butt-brushed” (Underhill’s term) makes it intolerable, driving them to forgo the concealer despite their need for it.

If you relied on data warehousing and data mining alone, you’d have adjusted your inventory planning to stock less concealer. This would have reduced waste – a good thing to do.

But by going into the store and watching shoppers actually shopping, you rely on observation – a more powerful tool than inference. You’d then know to move the concealer to a higher shelf, increasing sales instead of fine-tuning inventory. That’s much better.

There’s a double-barrelled lesson here for IT. The first is to recognize when an IT solution is incomplete. To take another retail example, I once discussed the possibility of setting up “register-pop” in a client’s stores. Similar to the familiar screen-pop used in CTI (computer-telephony integration) -enabled call centres, which helps telephone agents interact more effectively with callers, register-pop would identify shoppers at the register and provide information that could help store clerks up-sell.

Ultimately, we decided against this approach. Although technically feasible, it would have provided just-too-late information. By the time customers reach the register they aren’t shoppers anymore. They’re buyers – they have what they want, have already waited in line, and don’t want to go back into the store for more stuff.

Some advanced retailers are looking into an alternative that identifies shoppers before they reach the register. It relies on an advanced face-recognition technology called the Sales Associate. By giving the associate a wireless PDA connected to a customer database, they hope to make their sales force more effective in helping regular customers.

That’s the first lesson – that IT solutions are often incomplete. The second lesson is more universal: Infer when you must, but watch when you can. This is true whether the subject is merchan-dising or customer interface usability,

and whether or not a process design will work in your warehouse. Given a choice, don’t guess, don’t assume, don’t even ask. Position yourself unobtrusively and … observe.