Today I had a discussion with a consulting firm about their need to deliver operational BI to their clients, and the question came up as to whether we supported ad-hoc BI and query for end-users. The question was framed within the context of the traditional approach of writing an ad-hoc query and report, but I spent some time explaining how a much more interactive approach was a better answer to this requirement. Here is my rationale:
Typically someone wants to perform an ad-hoc query because they are trying to answer a specific question or obtain information to enable them to make a specific business decision. Given this as the backdrop which approach works best:
- Traditional query and report
- Think about the information you will need to answer the question;
- Specify a set of parameters and run a query;
- Get the result set in tabular format;
- Format the result to provide min/max/avg /aggregate values etc;
- Graph the result to see the historical trend etc; OR
- Interact with a graphical dashboard
- Interactively apply one or more filters for dimensions/aspects and time period and immediately see result;
- Ask for analysis lines to show trend, forecast, variance, control lines, Pareto, histogram etc and immediately see result;
- Select a chart, bar, segment or data point of interest and drill into that by any relevant dimension or its drivers (components used to calculate it) and see result;
- If needed, switch to a tabular view and see min/max/avg/medians;
- Perform a what-if scenario on one or more data points and immediately see the scenario result;
I believe the second approach is far more conducive to the decision making flow as I can follow a natural path to obtain more information as I need it, rather than trying to think up-front about a query that will return all the information I might need to make the decision.
Instead of ad-hoc query support, I think the real requirement is interactive investigation. What are your thoughts?
{ 0 comments… add one now }