Being There Means Knowing Where
Customer satisfaction may not be well measured by stopwatches or statistics.Its easy to spend a lot of money chasing the wrong definition of system availability. If I were selling processors, Id tell you that uptime comes from clusters and CPU failover mechanisms. If I were selling storage, Id tell you that consistent response depends on storage area networks and speedy switching. In customer-facing applications, however, the perception of high availability may be almost at right angles to the reality. If that expensive high-availability hardware is up and running, but the inventory systems arent well coupled with the order-taking processes, then customers may waste time ordering items that cant be shipped until after theyre needed. When a customer has crossed off an item on a list, only to get an e-mail four days later that says the item isnt available after all, thats not just a lost sale: its probably a lost customer as well, perhaps with multiplier effects due to negative word of mouth. "System down, please try again in ten minutes," might be bad, but isnt it better than, "Here we are, welcome to amateur hour"?
Conversely, a less gold-plated online ordering system might actually be down, but the Web site might still be able to offer the customer access to an online catalog combined with a one-click facility for e-mail inquiries and a promise of next-day reply. Properly implemented, a system like this need not be labor-intensive: on average, it could easily cost much less than adding more nines to your 99.99 percent availability, and it could actually yield higher customer satisfaction than a system thats always up but doesnt actually answer customers questions.