The VC in Vacation Land writes about real-time data. While I agree with the benefits he lists, it's important not to get too caught up in the hype put out by some business intelligence vendors. Yes, real-time is good, but not in all cases. For managing operations where minutes or even seconds can make a huge difference in scheduling, congestion, or labor allocation, it's a must. We have applications that do just that. But there are many situations where less current data is just as useful and there are definite costs associated with using real-time data.

We've been implementing an Inventory Visibility application and for one of our customers, updating the data once a day is good enough. In that case real-time would require us to access many different transactional applications on the fly which could impact performance for either the users of our dashboard or the transactional processing or both. By offloading the data into a separate data store, we control the data model so we can create reports and dashboard widgets very quickly. We can add indexes or redesign tables for performance which would be impossible if we were going against the transactional systems real-time. We have the option of refreshing more often if the business need changes. If we eventually need real-time, we can modify the extraction process to run continuously (possibly with some performance tweaks), without having to change anything on the front-end. And as a bonus, we can store the data we collect indefinitely, allowing us to do historical reporting and trending.

The bottom line is that you should evaluate just how real-time the data needs to be and weigh the costs against the benefit.