If you build a good site, users may come; but if they only visit once, you lose . This is one of the reasons why raw "hit rates" are discredited as a measure of site success: you can build seemingly impressive traffic volume by spiking your pages with search engine bait or by spending liberally on banner ads on other sites. You gain zero value from people who visit one or two pages and turn away in disgust when they discover that your site is not really about the topic they searched for or that you don't fulfill the promise of your ads. Site tourists crank up your hit count but do nothing for the long-term viability of your site. Repeat users are satisfied customers and the way to build a site.
Strong confirmation of the value of loyal users came from a recent study of Internet shoppers by Binary Compass Enterprises. The study found that new users at a merchant site spent an average of $127 per purchase, while repeat users spent almost twice as much, with an average of $251 . In this study, on-time delivery was a key predictor of customer loyalty, with 96 percent of shoppers who received their order on time saying that they were "likely" or "very likely" to shop at the site again. Delayed deliveries reduced the percentage of customers likely to patronize a site again to 60 percent.
If Soft Drink Sales Worked Like the Internet
When somebody tried a new soft drink and liked it, they would not simply tell five other people at their lunch table; instead, they would shout the news to a crowd of millions of people with an interest in soft drinks. Any of them who felt intrigued by the new drink could say "sure, let me try one" and would receive a sample can by FedEx twenty seconds later. If they liked the sample, they would be assured of finding ample supplies of the new drink on the shelves of every supermarket and convenience store in the world the next day.
If soft drink sales worked like the Internet, there is no doubt that new brands would be launched much more frequently and that consumers would change brand loyalties more than they currently do.
Unfortunately, speedy delivery is not all it takes to build loyalty, though traditional qualities of good customer service obviously should be retained on the Web. Web users are inherently quite suspicious and the nature of the Web encourages them to constantly visit competing sites: either because they follow a hyperlink to a recommended page or because they hear about a new service in a discussion group, newsgroup, mailing list, or chat room. It is so easy to sample new sites on the Web that users like doing so, and the communicative fabric of the Internet disseminates news and user reactions around the world in minimal time.
In the early days, Internet users were notoriously fickle and moved to new products at the drop of a hat. The Web has seen four stages of browser dominance:
- CERN line-mode browsing and other text-only interfaces (1991 and 1992)
- Mosaic (1993 and 1994)
- Netscape (1995 and 1996)
- Shared dominance of multiple browsers (1997)
The move from stage 1 to stage 2 and the move from stage 2 to stage 3 both happened extremely quickly. In both cases, the new browser ate more than twenty percent market share from the old browser every month, and an almost complete transfer of users happened in less than half a year. In other words, the old browser commanded very little loyalty.
The current switch to a multi-browser environment is happening much more slowly, with estimates being that Netscape only loses 1.2 percent market share per month. Even though 1.2 percent per month is still much faster than market share changes happen in the off-line world, the slower pace is an indication that Web users are becoming more set in their ways and less likely to move to the latest gadget. The implication is that there is some hope of building a loyal user base that does not evaporate every time a competing site opens up. Even so, loyalty is still harder to get on the Web than in the real world: for example, a convenient location does not matter, and somebody who sets up a server somewhere in South Dakota can get word-of-mouth referrals from your former users world-wide in a matter of days.
The classic way to increase loyalty on the Web is to have fresh content that changes on a predictable basis. Many sites even offer users an option to sign up for an email newsletter that reminds them when new material is available. Such update notifications are most valuable for sites that change at most once per week: any more frequent updates and the email becomes an annoyance for many users who will ask to be taken off the distribution list. It is invaluable to have users on a mailing list since they are almost guaranteed to keep visiting your site as long as you are allowed to remind them of its existence. Thus it may be better to reserve the update email for major events if a site changes too frequently. Alternatively, offer two mailing lists: one with daily notices for people who don't get enough email and another for the rest of us.
Getting users to bookmark your site is almost as good as getting them to voluntarily receive your email. Once you have a bookmark, you almost have that user for life: though data is sparse, current evidence suggests that users only rarely remove sites from their bookmark list. In the long term, I would expect browser vendors to improve bookmark management and offer features to, say, remove or downgrade bookmarks that have not been visited for ages. So far, though, bookmarks are forever. There is not much you can do to get users to bookmark your site, except making it possible to do so: no URL-eating frames , and no weird one-time-only links that do not work for subsequent visits. You can also create landmark pages that will be useful for repeat visits and thus tempt users to create bookmarks: for example, a page listing all the columns in a series, a page listing current best-sellers, or a page listing new products.
If you insist on user registration, passwords, or other cumbersome impediments to pure hyperlinking, you must reduce navigational overhead as much as possible. Users are unlikely to revisit a site often if they have to pass a Berlin Wall's worth of border guards every time. Use a cookie rather than an explicit log-in sequence to track users unless your site has high security requirements. If log-in is needed, use a standard log-in screen: in many cases, users will then be able to have the browser remember the userid and password for them. Usability studies show that users almost never remember their passwords, so they typically write them down anyway.
Loyalty is enhanced if the site becomes more valuable to the user at each visit. The simplest approach is a frequent-browser points , but it is also possible to make the site itself adapt to the user's needs and preferences over time. After several visits, the site will fit the user's needs so well that the user will be reluctant to move to another site that cannot offer the same service initially. The most traditional way to achieve this goal is to encourage the user to enter a user profile : for example, a portfolio listing on a stock site to allow one-click valuations, or keywords of interest on a news site to generate a personalized front page with the news of most interest to that user. The downside of these common user profiles is that they require a large amount of up-front work by the user before any pay-off can be enjoyed. Therefore, users will only establish user profiles on sites that have already gathered their loyalty.
Personalization works best if the site can gather information about users unobtrusively. For example, Amazon.com asks users to enter a complete shipping address for their first order, but if they order again, they will discover that the server has saved their address, making their second transaction much faster. Without any overhead, users enjoy faster processing if they keep ordering from Amazon.com than if they try out new online bookstores. As another example of unobtrusive personalization, search engines could build up databases of user preferences by tracking which of the search result links a user actually clicked on. After collecting this data for a while, the search engine would know what kinds of links the user preferred and it could give such links higher ratings in subsequent searches.
See commentary by a reader who fears that affiliates' programs with referral fees will reduce the quality of small sites' content and make them less motivated to serve users and promote loyalty.