Real-time Business Intelligence Using Analysis Services (Microsoft TechNet)
The goal of business intelligence (BI) systems is to enable better, more informed, and faster decisions. Hence any discussion of real-time business intelligence must be in the context of how close to real time the information must be to support those decisions. What constitutes real-time information can vary widely for different business activities, even within a single enterprise. For example, considering a chain of retail outlets:
- An analyst using historical data as input to the sales forecast for the next period would likely only need information as of the last month.
- A marketing manager evaluating the success of a campaign, and responsible for deciding how long the campaign should run, would need much more timely information, probably no more than one day old.
- A store manager might be making frequent decisions during the day; for example, deciding when to put a perishable item on sale. The manager would need information that was certainly no more than a few hours out of date.
Many times a combination of up-to-date and historical information is necessary. For example, while it may be useful for the store manager to know the number of sales of a particular item so far this morning, it is more useful if he can put that information in the context of the average number sales of that same item on the same day of the week over the last year.
Reports can include data from different systems. In our retail example, the data for today’s store sales would be taken from the source systems that are updated immediately at the time of sale, which would be combined with historical data in an enterprise data warehouse (DW). However, this means we no longer have "one version of the truth." When we compare data that is drawn from two sources, we must exercise great care to make sure that we apply the same business policies that were enforced during the load of the data warehouse. These include the consolidations that were applied, the treatment of discounts, and the approach to currency conversions. This means that the business rules embodied in the DW load process must also be embodied in the reports or applications that combine data from the two systems. This can result in a less agile and more error-prone system. This could lead to the making of decisions that are based on incorrect information.
Ideally, we want a single source that produces data that is close enough to real time to support all decision makers in the different scenarios. So what are the barriers that prevent business intelligence (BI) systems from supplying truly real-time data?