Several impressions and reflections on the last Hyperion conference ever.
The good news is that some of the best Hyperion folks on the planet are still with the company. That is to say some stellar Essbase product managers and developers are on the team and not going away. That includes guys like Marciante, Nader, Hite and Tai as well as very capable new folks I've never seen before. The even better news is that it's Essbase again. Not Hyperion Analysis Services Enterprise 9X or some huge mouthful. So when it gets right down to it, the Essbase we know and love is there in all its beauty.
The bad news is that I didn't have time to sit down with the Essbase guys and talk about futures. It was too crowded. But I did see some slides that said 'Trickle-Feed' and some that said 100s of Gigabytes OK. Since I have heard last year that the aim was for ASO to handle a full set of BSO functionality, eventually, I will have faith. So there is no doubt that Essbase has a future.
In fact, I think the case will be made abundantly clear going forward that the Essbase-centric paradigm of OLAP development will continue into the future. It is the platform for the future of BI. The Oracle-Hyperion people know exactly where the strengths and weaknesses of OLAP are, but Essbase still has got the advantage. That advantage depends on the combined company's ability to reposition the meaning of BI.
Yeah I said it was murky, but that's because the presence or absence of Essbase is not necessary to make sense of the application stack. It's not until you get into the details of how capable these applications are going to be at the individual level that you begin to recognize how key Essbase is to them. That's why Oracle is challenged to present BI in a whole new context.
The New BI Paradigm
People have been saying 'drill down' and 'slice and dice' for two decades now. And you would think that they would have gotten tired of that. Oracle is going to show them how retarded that is with their new technologies going forward. The reason is that now that System 9 is the big dog on the block, the APIs to Essbase are going to support a class of workflows that have never been so integrated before.
What the Hyperion engineers know, better than perhaps anyone on the planet, is how analytic workflows run. They have been doing complex OLAP for many years. They understand the cyclical nature of OLAP queries and iterative analysis, planning, goal-setting and review. Now this is what they are calling the BI paradigm, and what is new and exciting is that they are going to be exposing more of the data mining features of Essbase (that have been around for years) into the BI paradigm.
Obviously Essbase has always been able to handle multiple scenarios of data. There isn't one real application that we deliver that doesn't have at least three. What's new is that they are going to bring forecasting into the foreground with tools that will allow end users to drill down into a set of products (for example) and mine the past data, as well as the past generated data and automate the process of selecting heuristics, employ them and expose those heuristics for adjustment.
This means in Hyperion Planning, it will automatically fit curves at every level and generate a base forecast for you. It will crank those through and allow you to select for factors that effect the outcome, and recycle. This is way more intense than "take last year's sales number and add 10% and spread it". This is more like figure out which products have the most affinity with this new product we have and predict its sales, calculate the profit and forecast the profitability. Now take into consideration my promotions budget, and reforecast sales volume. Now what's the profitability? Now lets cut costs here and reallocate to tweak the price against this demand curve. What's profitability now?
That's the new paradigm of BI. It's not just reporting, it's smart extrapolation of financial and operational history and application of data mining for the individual user forecasting. It is a revolution just waiting to happen.
One of the comments that struck me in elaborating details of the merger was that Oracle was impressed with the amount of work that Hyperion produced per capita. As a relatively small company, the extent to which Hyperion absolutely dominated BPM was stunning to the Oracle folks. In addition, I've heard tell that they are staffing up their pre-sales and post-sales consulting groups. So all that points to more reasons to believe that as a wholly owned subsidiary, Hyperion will not lose, but possibly gain headcount within Oracle.
I've been in this business for a long time and even I had a hard time following the presentation by the guy from Decisioneering. What I can tell you is that a whole new era of risk management is coming to be integrated into the planning cycle. That is because unlike the products I've seen before with SAS and SPSS, this stuff is totally integrated with Excel. Excel makes the world go around, and Essbase connects that world with the universe of business data. But I just learned how to do a 10,000 case Monte Carlo Simulation and a 10 factor sensitivity analysis in under an hour. I saw curve fitting and confidence intervals put in place and I saw a 4 year forecast built in front of my eyes that was simply brilliant. And guess what? They did it at Ludicrous Speed.
By far the coolest thing that I learned at the conference, aside from the fact that I've still got my presentation chops, is that the guys at Tellme are using MapReduce to get massive data into Essbase. What does that mean? It means that I now have a framework for integrating VLDB data into an analytic platform.
Tellme basically generates 100GB of data daily. Huh? What? That's what I said. 100GB daily. it comes from processing 5 million telephone calls which generate on the order of 150 million detail records, PER DAY! They've got something like 30 MySQL master / slave arrangements in a two-tier MapReduce network. They issue a query through this very cheap, open source network of PCs in about 4 minutes. That's right, in 4 minutes they have processed all of the aggregate data they need into an ASO cube. That, my friends is scalability.
The second coolest thing at this conference was SmartSpace. Basically, this is the future look and feel of BI. All I can say about it is this: Forget portals. Forget Web 2.0. It's all about the widgets. SmartSpace is, make no mistake about it, a killer app for Windows Vista. I could not take my eyes off of the screen. It was absolutely gorgeous. It's a single signon, intergrated IM, multi-canvas workspace for System 9 that deploys sharable widgets for every kind of display and control available to the Hyperion stack. It turns a PC for BI what Photoshop does for a Mac. There are auto-hiding toolbars, floating controls like a calc script initialization widget, for example. I didn't think anything would make me take a second look at all this again, but this is the front-end of the future. Oh by the way, it also works like a thumbdrive. You 'mount' a SmartSpace and double click and you get the hierarchy of System 9 Workspace right in the Windows explorer. I'm in the Beta, buddy.