Eureka Blog

Looking back on my 15 years of software experience, it's amazing to me how pervasive the need for analytics has been. Once you have a piece of operational software even close to working, people start asking:

Even outside the software products, we were always internally asking the same sort of questions:

• What is our acceptance rate on new hires from top tier schools?
• What's been our monthly outsourcing spend over the past half?
• How many defect are we finding per day? Who's finding the most?

You'd think with such a pervasive need and more than 10 years of a booming internet the available analytic options wouldn't suck so much.

As an end-user, there are two important criteria that stand out when looking at analytic tools.

First, how easy is it to explore your data to find the answers your looking for? Part of it has to do with the capabilities of the tool - how easy is it to drill down, drill through, backtrack, etc. Another important (and overlooked) aspect is how fast the system is. Regardless of the capabilities of the tool, if it takes minutes to get an answer, you're not going to be able to explore your data. (I'd say close to 90% of the Fortune 500/Global 2000 deployments I've seen do not have "exploration" capabilities, precisely for this reason).

The second criteria is how easy it is to share your discoveries out once you've found them, which screams for a web paradigm. Rarely does someone do some analysis, look at the answer, and then delete everything. The whole point is to share insights with other people. You want the information to be easy to disseminate, and widely (but securely) available.

The problem has been until recently these two criteria have been at odds. Web-based tools are a pale shadow of what you can do on the desktop. Most online analytic tools are bastard children of previous desktop applications. They look worse, interaction is horrible, and scale to large results poorly. AJAX based technology has only been marginally helpful. At one previous job we OEM'd a DHTML, web-based analytic tool that would come to a standstill displaying more than a couple hundred data points. They suggested slipping in a filter between the data and the user, and only providing a 'sample' of the data set, to keep the number of data items down to a something the tool could handle.

Besides web-based tools, Excel is the other choice, and, by far, the most successful BI/analytic tool ever. It's easy to get data in and manage it without needing an IT staff. It's relatively good at exploration - fairly easy to build views, and usually performance is decent. (Although it's not well suited for backtracking during exploration). Excel does have one major problem for analytics:

It's horrifically bad for sharing and distribution. In Excel, sharing usually means making a copy and emailing it. Not only a copy of the view(s), but worse, a copy of the data. Sadly, some excel-sized problems get moved into a data warehouse simply because data ownership becomes a huge problem - you have multiple versions of the data floating around.

Here's a common excel scenario: You have a fairly large interesting data set, portions of which are applicable to a wide range of people. You create 20 different views for each user set, 1 worksheet per view (pivot, chart). You email the massive excel file to everyone interested. Keep mailing it out every time the data changes. God forbid anybody else makes other changes/corrections and you have to track them down and reconcile.

Another scenario: You slice and dice your way to an interesting view of data - which gives you another idea of where you want to explore. You can either: undo back to a good starting place and lose your current interesting view. Create another worksheet, copy paste, and then backup through undo.

With the advent of online spreadsheets, centralized management of your moderately sized data sets is now extremely easy. There is absolutely NO REASON to go stick it in a data warehouse. And now that you have your data centralized online, you want to analyze it and distribute it just as easily.

We think we can help, and we've just started scratching the itch.