Applications:

Sand's Nucleus Optimized for Analytics

You know your data warehouse. You like your data warehouse. OK, maybe you don't really like it, but it's still a key part of your business, right? But are you really maximizing its value?

Sand Technology hopes that most companies will answer that question with a resounding no, because the Montreal-based software maker thinks it has a better answer.

Sand Technology's flagship product is called Nucleus. To an end user, Nucleus looks pretty much like a standard relational database. But behind the scenes, Nucleus is optimized for analytics.

How does it work? Nucleus takes tables of data and stores them as numbered columns. (The value of column "Name" in Row 3 might be "Bob," for instance.) Nucleus then removes all redundant data by replacing values with tokens. For instance, a given customer database may include Boise 200 times and Los Angeles 1,500 times -- Nucleus just stores the city names once and assigns all occurrences the same number.

This storage technique has a couple interesting side effects. First, it saves disk space, and it reduces memory requirements when running queries. Second, it eliminates the need to search for text strings. Instead, the computer simply -- and very quickly -- scans rows for a given number. That means queries once too complex to consider become reasonable, even simple.

Sand's execs know they have some work ahead, of course. The company's technology works best on very large, complex data sets. As such, they're targeting the government, medical research and telco space in North America first. The company's first big North American win -- Montreal-based Sherbrooke University Hospital -- plans to use Nucleus to mine 12 years worth of records for more than 200,000 patients to help determine best practices for diagnoses. And, of course, many of Sand's potential customers already have data warehouses in place. (Sand's European market president, Duncan Painter, acknowledges that the company will need to sell Nucleus as an adjunct to entrenched systems in those cases.)

To make things a bit smoother (they hope) Sand comes in at a less-than lofty price point. A recent European project (the company already has more than 30 customers outside North America) to combine data from five different companies cost around $350,000. The larger Sherbrooke deal is listed at $2 million. Neither amount is crushingly large by data warehouse standards. Better yet, Sand will guarantee its price -- no creeping integration and configuration fees -- and it can generally get systems up and running in two or three months.

It's not an easy time to break into new markets, but Sand looks like it might have a better shot than most. If you'd like to find out more, visit www.sandtechnology.com.