The Internet is on its way to creating perhaps the most comprehensive customer-satisfaction survey ever.

With the Web, companies can track and analyze customer behavior, not just ask their opinions. It provides instant feedback that can be used to build relationships, or a two-way dialog, between customer and company. It also gives companies a quick way to do something about their ratings—there are countless on-line opportunities for companies to learn continually what their customers want, so they can refine their products and services.

Look at Vodac, South Africa’s fastest-growing cell-phone company. Vodac (www.vodac.co.za) counts more than a half-million happy customers because it has constantly refined its Web-based services, especially for business users. For instance, Vodac lets each user build a personal phone book of 100 numbers that can be dialed automatically. Even better, Vodac manages those numbers like a corporate voice-mail system. Users can send a single message to several numbers on the list or broadcast a message to groups within the list.

Because Vodac can closely monitor behavior on its Web site, it has been able to see just how each new capability is received—and just how many, or not so many, customers have become dissatisfied and moved on to a competitor.

What Vodac is doing is great news for all those companies that focus so heavily on satisfaction ratings, comparing one division with another, this year with last, and even using the ratings as a basis for annual bonuses. Vodac shows how to take a customer-satisfaction focus to the next level. That is crucial, because satisfaction will become even more important as more business is done on-line, where customers who have a problem can move to a competitor so easily.

As you think about how to use your Web site as a customer-satisfaction tool, here are two things to keep in mind:

COMPARE, COMPARE, COMPARE. It may seem obvious that the numbers assigned to customer-satisfaction ratings don’t mean anything in absolute terms. But how should they be analyzed? What should they be compared with? Companies get confused.

AT&T, for instance, started examining customer satisfaction more than a decade ago but initially found that scores apparently had no correlation whatsoever to market share, growth, or profit. A typical enigma went like this: Customer satisfaction for one product was reported at 97% in Pennsylvania and 78% in New York. Yet the product was losing market share in Pennsylvania and gaining share in New York. How could that be?

It isn’t that hard to explain. New Yorkers rated everybody lower than the more tolerant folks in Pennsylvania did. The comparison between New York and Pennsylvania simply wasn’t relevant. The real question was how AT&T’s ratings compared with competitors’. As it turned out, New Yorkers rated other phone companies’ products even lower than AT&T’s, while the company didn’t fare as well in Pennsylvania.

AT&T found that when relative customer satisfaction was measured, the correlation with business performance was remarkable. AT&T learned that even a small uptick in satisfaction, in comparison with competitors, was a leading indicator of market share growth four months in the future.

HONESTY IS THE BEST—ACTUALLY, THE ONLY—POLICY. Companies have to make sure that managing for a high customer-satisfaction rating doesn’t produce some unintended and distorted results. That is particularly true when ratings are tied to compensation.

A friend of ours recently bought a new Toyota from a dealership that advertised its very high satisfaction ratings. As our friend was signing the papers, the salesman explained that the dealership’s revenue and his personal compensation depended on how the customer rated his satisfaction. The salesman slid a satisfaction survey across the desk and asked (wink, wink) whether our friend hadn’t been very satisfied with the sales experience. The survey came with a postage-paid envelope, but the salesman encouraged the customer to fill out the survey right then, in front of him. After all, he could save himself the trouble of finding a mailbox later, and if he really was very satisfied, why not?

Don’t let yourself get caught up in that salesman’s game. Satisfaction ratings may be artifacts that can be manipulated, but preferences are for real.

Peppers and Rogers are partners in Peppers & Rogers Group, which provides strategic consulting to companies on how to build collaborative relationships with individual customers. Its Web site is www.1to1.com.