Thursday, November 06, 2008

LucidEra and Birst Blaze New Trails for On-Demand BI

I spent a few minutes last week on the Web sites of about eight or nine on-demand business intelligence vendors, and within a few days received emails from two of them ostensibly asking about much earlier visits where I must have registered with my email address. Given my current obsession with demand generation systems, I’m pretty sure this was no coincidence: they had deposited a cookie linked to my email address during the earlier visit, and used this address to react when I returned. I suppose I should admire this as good marketing, although the disingenuousness of the messages was a bit disturbing. I suppose they felt that was better than the creepy feeling I might get if they said they knew I had visited. [Postscript: I later spoke with one of the salespeople, who sadly assured me it was a total coincidence. She only wished her firm could react so effectively.]

(I was about to coin the word “disingenuity” to mean something that is ingeniously disingenuous [i.e., cleverly deceptive], but see that the dictionary already lists it as a synonym for disingenuousness. Pity. )

Whatever. The reason I was looking at the on-demand BI sites was I’d spoken recently with two vendors in the field and wanted to get some context. One of the two was LucidEra , which was giving me an update since my post about them in July.

They’re doing quite well, thanks, and most excited about a new services offering they call a “pipeline healthcheck”. This is a standardized analysis of a company’s sales pipeline to find actionable insights. LucidEra says it has been tremendously successful in demonstrating the value of their system and thus closing sales. Apparently, many marketers never learned how to analyze the information buried within their sales automation systems, simply because it wasn’t available back when they were being trained. So doing it for them, and helping them learn to do it for themselves, adds great value.

This reinforced one of my few really profound insights into the software business, which is that the marketing software vendors who succeed have been the ones who provide extensive services to help their clients gain value from their systems. (Well, I think it's profound.) Interestingly, when I told LucidEra I have recently been applying this insight to demand generation vendors, they said they had recently switched to a new demand generation vendor and—this is the interesting part—found the new system was so much simpler to use that very little vendor support was necessary. That’s an interesting tidbit, although it doesn’t necessary confirm my service-is-essential thesis. Perhaps it needs a corollary of some sort when the applications are obvious or the users are already trained. Facts can be so pesky.

The other vendor on my mind was Birst. I actually spoke to them back in early September, but their product announcement was under embargo until September 30 and in any case I’ve been focused since then on the demand generation guide (have I mentioned http://www.raabguide.com/ yet today?) I’m glad to get back to Birst, though, because I was quite intrigued by what they showed me. Basically they claim to have fully automated the entire business intelligence implementation process: loading the data, designing the warehouse, identifying interesting information, and creating dashboards to display the results.

I’ll admit to being skeptical of how well they can do this, but the company’s managers have some excellent credentials and Birst itself is a project of a Success Metrics, which has been providing Web-based opportunity discovery to insurance and pharmaceutical sales forces since 2006. They offered me an online workspace to play with the tool, but I haven’t had time to take them up on it. (I think their Web site makes that same offer to anyone.)

I did spend a few minutes playing with a prebuilt demo on the Web site: it’s a reasonable user interface for ad hoc analysis and building dashboard reports. There was a lag of up to five seconds between each click when I was working with the data, which would quickly get annoying if I were trying to do real work. Part of the lag may be caused by the underlying technology, which generates relational OLAP cubes on the fly in response to user queries. But it also appears the system uses a traditional Web interface, which redraws the screen after each click, rather than AJAX and similar technologies which provide a smoother, faster user experience.

I don’t want to dwell on the Birst user interface, partly because I haven’t tested it thoroughly and partly because you can judge it for yourself, but mostly because their more important claim is the automated implementation. As I said last March, I think the labor involved with building the system is the biggest obstacle to on-demand BI, so Birst’s claim to have solved this is the real news.

It would take some serious testing to assess how good a job Birst’s automated systems can really do. Still, the system can be useful even if it’s not perfect and it will presumably improve over time. So if you’re thinking about on-demand business intelligence, either for a specific purpose or just to better understand what’s possible, Birst is certainly worth a look.

Incidentally, my quick scan of other on-demand business intelligence vendors (Autometrics, BlinkLogic, Good Data, oco, OnDemandIQ, and PivotLink) showed that only oco made a similar claim about having automated the implementation process.

On the other hand, Good Data, PivotLink LucidEra and possibly oco are using in-memory or columnar databases (PivotLink’s is in-memory and columnar: they win). In theory these should give quicker response than Birst’s on-the-fly OLAP cubes, although actual performance depends on the implementation details. (Speaking of experience, Birst’s database technology has been running at Success Metrics for several years, and has scaled to the terabyte range. I don’t know what scales the other vendors have reached.) It also seems to me that in-memory and columnar databases should be particularly compatible with automated implementation because their simpler structures and greater efficiency make them more forgiving than conventional databases if the automated design is less than optimal. But no one in this particular group of vendors seems to have put the two together.

I don’t know when I’ll have time to give all these other vendors the attention they deserve. But based on what I’ve heard from LucidEra and Birst, and seen on the other vendors’ Web sites, I’m more optimistic about the potential of on-demand business intelligence than I was back in March.

No comments: