Q&A: Getting Back to OLAP Basics
Almost one year ago to the day, Applix jettisoned its CRM assets to refocus on its TM1 OLAP engine, the original source of the company's growth. We look back on that move, and ahead to the company's interest in business performance management.
Last year former OLAP vendor Applix Inc. got back to the basics.
Almost one year ago to the day, Applix divested itself of its CRM assets and announced its rededication to a technology—the respected TM1 OLAP engine—that provided much of its growth in the first place.
Since then, Applix has been as good as its word, introducing a new version of TM1 that lets users read and change data from the context of either an Excel spreadsheet or a Web browser. We spoke last week with Dave Menninger, vice president of marketing for Applix, who discussed how this once and future OLAP high-flyer also has its sights set on the business performance management market.
It was an eventful 2003 for Applix, as you sold your CRM solution stack [to Platinum Equity] and rededicated yourself to your underlying OLAP server and business performance management technology. Can you talk about some of the factors that influenced your decision to divest yourself of your CRM investment?
The CRM market had become extremely competitive. Microsoft had not yet entered the market, but the big CRM players were all moving to the mid-market, and we were very much a mid-market player. We had seen our revenues become fairly flat (in fact, decline), and you have a stable of horses, and you have some that are running better than others, and you make the decision to focus on what’s better for the shareholders. An important point is that in the CRM world, our technology was good, but I wouldn’t consider it unique.
One potential benefit of this move was to make Applix an attractive choice for analytics vendors who want to OEM a third-party OLAP engine but who would otherwise be in competition with Applix on the CRM side of things. Have you seen any interest from vendors on this front?
We’ve had a similar effect, but not specifically with CRM vendors. Since the announcement [that Applix was quitting the CRM business], we have been successful at raising the visibility of the company and identifying the key strengths of TM1. We’re not Hyperion, IBM, or Microsoft, who many people compete with, so we weren’t encountering that [disinclination to OEM TM1 because of competitive concerns] as much. Interest tends to be so far more from the infrastructure players, and we are pursuing aggressively the ERP vendors as another potential source of licensing.
You announced a major update of your TM1 OLAP server last year. Could you talk about what the new version of TM1 brings to the table?
We introduced the product in September, and customers have been very receptive to it. The new TM1 Web [a Web spreadsheet feature that integrates with Microsoft Excel and supports read-write access to data stored in TM1] is an improvement over a different Web mechanism we had that was much less rich, so it had the capabilities of displaying both tabular and chart data, but not the rich integration with Excel.
So with TM1 Web, we now have Excel worksheets that reference TM1 data and literally publish that information to the Web. It’s literally a menu [in the Excel client] that you can pick to publish data to the Web. You can also edit data from a Web browser, exactly as you would in Excel and have the changes recorded [on the TM1 server].
We’ve labeled what we do [with TM1] BI Share Technology, which is a notion that these people are comfortable on Excel and we’re giving them an environment to share this BI information. The simplest way to describe it is that what you can do in Excel you can do over the Web. It’s easier and sometimes provides better performance if you’ve got a Web-based environment as opposed to a client-server environment, but you can choose between both.
OLAP benchmarks tend to focus overwhelmingly on query performance, but some industry watchers argue that load-times factor just as much into the equation. How does TM1’s architecture help with this?
The most important thing to understand about TM1 is that it’s memory based. If you were to compare it with Hyperion Essbase or Microsoft Analysis Services, they are both products that assume or rely on the fact that you’re going to have some large amount of data stored on disk, and how do you optimize getting that data into and out of memory, so it has certain kinds of characteristics that it’s good for.
Our approach is to do everything in memory, and as a result we optimize for recalculating the data, so we can quickly recalculate and show the results of that change. You can literally do this interactively: You change data on the screen and see the results in seconds, or—if it’s a really large application—within minutes. In fact, the most recent OLAP survey [conducted by OLAP authority Nigel Pendse in conjunction with Survey.com]—OLAP Survey 3—ranked TM1 the highest or best performing in terms of query times and data load, pre-calculation times. So the amount of time it takes to get data into TM1 and the amount of time it takes to get data out of TM1 is unequalled.
Wouldn’t a memory-based environment pose other limitations, however, in terms of the size and cost of the systems that can be supported?
If it’s memory-based, you’d assume that there are scalability issues, but we’ve done a couple of things to overcome that. First of all, we’ve added a 64-bit environment, so if they encounter scalability problems [in the 32-bit address space], they have the option of using that. Most of our customers are able to fit very comfortably in the 3 GB Wintel 32-bit limits. I would estimate that roughly 90 percent of our customers use the 3 GB [limit].
Have you any customers that are using the 64-bit edition of TM1?
We have a customer called Pelephone. They’re a provider of cellular telephone service in Israel, they have about 1.5 million customers, and they take every call that every user makes every day and they analyze what that subscriber would be paying if they use[d] a competitor's rate plan. The idea is to target the ones that would be likely to churn or switch to a competitor, and make them an offer before they leave, so you can imagine why they’d need to go to the 64-bit version!
You’ve also been focusing pretty heavily on the business process management (BPM) side of things. Aside from the strength of the TM1 engine, what do you think bring to the table that helps to set you apart from other analytics vendors, say Cognos and Hyperion, who also compete there?
Clearly the Excel integration. Most other vendors treat Excel somewhat as an afterthought, with Hyperion probably being the closest to Exploiting Excel. But we really leverage Excel to the fullest, from an end-user perspective. That, I think, is an important differentiator. Our customers typically live in Excel.
Another difference is that we’ve been told that our logging capabilities are unequaled. Every change made to every data value everywhere in the system is logged in a log file, and I think that some of the other vendors can make the same claim, but we can selectively reverse any of those changes. So if you, in the process of auditing some of the changes, determine that changes were made that that shouldn’t have been made, you can literally highlight them and ask the server to reverse them.
This is especially important for Sarbanes-Oxley and other areas of compliance, because we make it auditable: Every change that’s made to the data is tracked, all of the presentation is capture in the spreadsheet, but all of the data and the logic as to how the data’s presented resides in the server.
Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.