High performance DW rule #5 Push the data to the users, don't make them pull it.

"Pull the string and it will follow wherever you wish. Push it, and it will go nowhere" --Dwight D. Eisenhower

Yeah, this absolutely doesn't work in the data world. It works everywhere else, so he was on to something, but no rule is a rule unless there's an exception.

One of the most interesting cost centers for BI applications is users pulling data. I don't blame them, if I were them and had no other choice, I'd do it too.  

Every day, every manager in my company fire up our given BI tool, goes through lord knows how many dashboards, and the result? 90%+ of these managers gain no value from this exercise because nothing changed on their teams, but they just contributed to my DWH costs. Every dashboard loaded is costs.

It's mind numbing.

Imagine you're a manager in a call center. You and your team are judged by ideal metrics: on time, backlog, volume, rate and quality of each customer engagement.

You come to work in the morning and scan various dashboards to see who's on time metric is exciting. At this point, you're manually scanning for anomolies. If everyone's doing what they've always been doing, there's no action to be taken. If Jeff suddenly has an on time metric for yesterday that's half a standard deviation above or below his norm, this is important. If it's low, you need to coach Jeff up to get him back on track. If it's high, and all of his other metrics are normal, we need to ask him what he did and coach the rest of the team up to do that. But most days, you find nothing actionable.

Of course, if the whole team is above or below normal, we need to talk to other teams to figure out what we did wrong, or teach other teams what we did right.

The interesting thing is that the data warehouse already knows the entire history of Jeff's on time metric, heck it knows the entire history of everyone's on time metrics, as well as aggregate team and division metrics. There's really no need for anyone to go through the process of hunting these things down. We can run this query once a day rather than the hundreds or thousands of times managers and individual contributors will run it trying to find the same thing.

So push the data. For an individual contributor, when their performance on a given metric breaks the norm either historically or compared to their peers, send an alert to them and their supervisor. Same is true of entire teams. For bonus points, log the anomaly and require the manager submit a record to the database explaining what happened. Very quickly, your DW costs will drop, your BI staff will have more time to work on revenue generating processes and the amazing part: your company will improve, continuously.

This will deliver immediate value to the company AND reduce costs, both platform and manpower.  And since everyone's data is in front of them all the time, if you mess up you'll get immediate feedback, increasing data quality.