Simon Thorneycroft and Jonathan Hodgson

Business Intelligence, KPIs, Data mining and associative algorithms

Business intelligence is a term that's getting alot of press these days, especially with Sql Server 2005 Analysis Services now released.

We have Key Performance Indicators (KPIs) which can be described as:

In business terminology, a key performance indicator (KPI) is a quantifiable measurement for gauging business success. A KPI is frequently evaluated over time. For example, the sales department of an organization may use monthly gross profit as a key performance indicator, but the human resources department of the same organization may use quarterly employee turnover. Each is an example of a KPI. Business executives frequently consume KPIs that are grouped together in a business scorecard to obtain a quick and accurate historical summary of business success.

These 'business scorecards' can be a collection of KPIs displayed on a portal such as SharePoint, as in this example of how Microsoft Does IT: Measuring IT Heath with the IT Scorecard. Lots more examples in this paper on using Sql Server 2005 Analysis Services for Solving Business Problems.

Data mining and data warehousing and OLAP can be quite new terms to most relational OLTP database developers but shouldn't be. Introduction to Sql Server 2005 Data Mining goes into the whys, hows and some of the models available.

I'm currently reading Data Mining with Sql Server 2005 which walks through the different mining algorithms and has some easy to understand examples like Movie!Click for associative mining.

"People who liked this, might also like this" - these types of results can sometimes be done using relational queries and good meta-data, for example if viewing a document on the Technology sector you might be interested in other documents in the same sector. But associative & clustering data mining can go further statistically basing results on what other users have also read for example.

On the subject of querying and 'information overload' a company called Information Mapping have an interesting product which can produce summaries of text - similar to the AutoSummarize functionality available in Microsoft Word but more advanced. From their example, it can take a conventional version of a memo:

And convert it to:

Also IBM's WebFountain shows that lots of research is happening in the search and natural language processing space, but has a way to go before it's mainstream. This example on how parsing statements for meta-data to improve searching: