Cubegeek

Business Language Search

I mentioned that I would cover some more of the detail of Kopke's technology direction. Well, here we go.
The first of the announced futures for additional features/ functions and tech in the Hyperion BI platform is:

Business Language Search
The tagline for this is 'easy as Google'. It's not Google, but apparently it doesn't have to be. What information I was able to make sense of is that there is, within the metadata of templates built for reporting, all of the keywords I need to get high reliability hits on English language queries. I basically asked Kopke if he had heard of or was scared of the Cognos - Google announcement and the answer was a resounding no. Of course what else would he say? Well he said that Hyperion had been there and done that with another search vendor and several years ago and had learned some lessons, and that these would be applied to the new stuff.

At R&D Central, this function was demonstrated in a fairly raw form, but the big slides at the opening session presentation were not simple smoke and mirrors. There is an ability for Hyperion stuff to execute OneBox style queries with thumbnails of analyses to the left of text describing the template. I'll see if I can get a screenshot for you. It looks good.

Tangentially, my old pal & mentor Dyke Henson who runs marketing over at SPSS is looking at search from a different perspective. The core at the center of these are ontologies, which I'm tripping on these days. So here's the deep dive:

SPSS is giant steps ahead of most folks when it comes to data mining and predictive analytics. They have a set of advanced tools that they are going to package around 'Predictive Analytics' and drink up the web. Then they'll use their advance mining techniques to find outliers in influence. So what had gone before with regards to attention in print media and banner ads is now going a level deeper with regard to say... link maps in the blogosphere which are subject specific.

The key to both of these are ontologies. As far as Hyperion is concerned, it's a done deal. They already know the universe of masterdata that is queriable, and the end result is already templated in terms of the objects which will be queried live when the user selects a thumbnail. So that search problem is not very big, and Hyperion will add value with item number two which I'll cover separately. Adding an ontology to this which is searchable shouldn't be too difficult, and metrics about who queries what will be readily available. In a search universe of an enterprise set of say 4000 concurrent users.. it's a peice of cake.

For Dyke, the problem is a bit more daunting because user-defined ontologies (oh did i not dumb that down - hierarchy of keywords for categorizing content) are fuzzy. You know how when you misspell a word at Gogle it ask 'did you mean Google?'. That problem. So when you are processing a huge segment of the web, or even top dogs in the blogosphere the various tags that people might use could be insufficient and poorly managed. So they have a little bit of a stickier problem, but I have confidence that they'll get a handle on it. Can I say that with a .42 assurance? Yeah, I guess.

Liveblogging Solutions 2006 - Day One - Part One

I have breathed a sigh of relief. Although Dresner was able to deftly sidestep the question of a Google collaboration, I have a good idea that Kopke is very well aware and is decidely interested in duplicating the functionality, if not the agreement that Cognos has made with Google vis a vis OneBox Enterprise.

The opening session this morning was an order of magnitude more confident than that of last year's at New Orleans. The acceptance rate of new customers for System 9 is, according to Sullivan, moving forward nicely. Dresner is brilliant in his strategic vision of the BPM revolution and quite honestly, here is the kind of merging of technology vision, practical application and straight communications that has finally eclipsed the term OLAP. Sorry Cruikshank, but you've finally been outdone. There's plenty to drilldown into with Dresner's talking points which I'll do under separate cover, but it works out nicely and jibes very well with my own experience.

Kopke's presentation showed 6 concise technology directions that extend the concepts of Dresner's vision, and I think I have been sold by demonstrated product realizations that the Hyperion development organization is quite capable of pulling it off. The very first bullet point was simplicity and they are determined to make this stuff and I quote "..as easy as Google".

What is clear in Hyperion's product direction is that they have thought long and hard way outside of the boxes of 'BI' and 'OLAP'. The technologies that they are ramping up demonstrate a deep understanding of how companies actually make decisions and they are creating tools in just about every space. Think of the Hyperion suite of applications and technologies as coalescing around the management decision workflow of organizations. Every place a manager leaves his signature on a decision that affects the way people work - if it can be measured, there will be a piece of Hyperion's software there. This is so much more concrete than an over extension of single large products into a jerrymandered platform. This is a swarm of technologies on a concrete platform, and Hyperion has the confidence to keep building more and more products into this swarm.

At the R&D Central, a gent named Chowduhry took the time to show off Hyperion's latest innovation in time series smarts. As you may know, I used to work for Pilot Software. The first thing we'd ask any customer was what about their periodicity of data. A good $10 word that. We would explain that we could handle, natively, any calendar that they used and that we were the only company that could. The first time I saw Essbase, it was the first question I asked. They said that a time dimenions (according to Codd) was nothing special and that I could build it with any grain that I wanted, which was true, but it didn't help me build it. So there has always been a bit of development noodling around time issues. Nothing we couldn't do, but sometimes painstaking.

No more. Hyperion has built the mother of all calendar wizards. I thought it was something when I saw the load rule builder that showed 6 ways to build dimensions. But wait until you see this wizard that can define your fiscal year 6 different ways. And that's just for fiscal calendars. Did you know about ISO 8601 calendars? Did you know the National Retail Federation has a standard for retail calendars? Hyperion knows and the wizard is built. Technically, this bad boy will use MDX and wizard-built attribute dimensions to manage time crosstabs very smartly. So you could do stuff like current year trimesters versus prior year trimesters on weekly data in 3 'dimensions'. It handles stuff like "my fiscal year begins on the second tuesday of every March and we do a 4-4-5". All obscure calendar issues are now wizarded into submission.

The Inmon-Kimball DW integration session was a bit heavy on theory and light on practice. I was falling asleep from the delicious salmon from lunch. I'm rather stunned that this is my 10th year in the business and there seems to be no end in sight for growth. All of the consulting partners have more than enough work just keeping up with demand and everybody here is eager and excited. Old faces are cycling through the industry and lots of new faces are showing up too.

Oh by the way. The new Hyperion logo is a big winner. Happy 25th! Also, I'm very curious as to what the MSFT and HYSL announcement is going to be tomorrow afternoon. I have no clue, but I think it will have to do with ProClarity's universal MDX support as well as Yukon's data pumping capabilities. Basically, I think MSAS cubes will be able to talk more or less directly with Essbase and vice-versa. This should be interesting.

Stay Tuned.

OneBox: Ignore at Your Peril

Google jointly announced OneBox today with Cognos, Oracle, SAS and others.

We launched a new version of the Google Search Appliance today with a feature called Google OneBox for Enterprise. Ever search on Google for a stock ticker or the weather and get the information right in the search results? Well, Google OneBox for Enterprise works the same way - giving you access to information from business applications such as your contact database, calendar, CRM or BI system. Check out Dave Girouard'’s blog post for more on how this works.

We're thrilled with the impressive list of partners (Cognos, Oracle, Cisco, Salesforce.com, SAS and many others) who have already created OneBox modules. Take a peek at this video to hear some of our partners talk about why they think a Google search box is a great way to reach their applications.

This is the day I've anticipated and feared. It is the beginning of the end for the BI world as we know it. Why? Because it fundamentally changes the way people query for data. Let's face it. It takes more brains to know how to run a visual query paradigm than a text search box. Everybody is better at web surfing that navigating query tools. Nobody ever, ever, ever got one minute of training to use Google. Everybody demands it for enterprise applications including BPM, CRM, BI, DW.. you name it.

At any rate, I'm going to start investigating the headache this new wrinkle is going to cause my career. But here's my assessment so far with regard to Essbase based upon a couple PDFs I read from Cognos.

it makes sense that the Google Appliance is more efficient at processing queries in terms of throughput than a database server on generic hardware. The Google Appliance is basically Britton Lee for text. So if by generating the contextual stuff for BI content, the GA can pitch stuff faster than, say Cognos Powerplay, then half of the BI tools market is under threat. And you gotta admit that Ajax stuff can be done right. So now you've got a zero-footprint, zero-training mechanism to deliver flat, keyword indexed BI data. That means you need a smart data generator pre-guessing what might be queried to be chugging along in the background as data is generated from various sources. (It also might mean you have to deal with authority of multiple sources...but more onthat later).

Essbase already has a way to optimize slices of multidimensional data based upon queries. The question is whether the GA is querying the database directly or just indexing data already dumped out by the database. It seems to me that the Google API is capable of doing both simultaneously and since Essbase can respond very rapidly, why not do both?

So basically, this is a middle tier problem for Microsoft and Hyperion. It very much reminds me of the system I hacked together for 3Com with AvantGo and Hyperion Spiderman. So this reminds me that if there is a clever way to embed masterdata keywords into Hyperion Reports HTML rendering as well as some hard coded linking (whew, echoes of Essbase Web Gateway), that HYSL could be all good. The question is whether or not they are paying attention.

They had better be.

Struggling US Auto Suppliers Ignoring up to $7.6 Billion in Cash

Tied up in Excess Working Capital, Hackett-REL Research Shows

Hard-pressed companies face even greater liquidity pressure in 2006

ATLANTA, April 11, 2006 – In the face of bankruptcy filings, lagging sales and rising raw-material prices, the 20 largest US automotive suppliers are nevertheless ignoring up to $7.6 billion in cash opportunity, according to new research from Hackett-REL, part of The Hackett Group Answerthink, Inc. (NASDAQ: ANSR). The opportunity comes in the form of excess working capital tied up in invoices paid late by customers, suppliers being paid too early and inventory lying unsold on warehouse shelves.

Results of the Hackett-REL benchmark study and comparative analysis, “Fine- Tuning for Optimum Cash and Cost Performance,” also determined that the top 17 European automotive suppliers are similarly overlooking up to $9 billion in cash, which is frozen in excess working capital. On average, US auto suppliers significantly outperform their European counterparts, showing 37% lower net working capital ratios. But some of this gap, and the strong working capital performance by several US auto suppliers that are already in or near bankruptcy, is likely to have been driven by special assistance programs from US automakers, and may disappear in 2006.

Two Seconds Flat

I've been thinking about explaining the basics of BI to noobs as well as senior managers whose companies have yet to exploit the economies. In answering an imaginary question of where the industry will be in 10 years (actually the iQuestion was 'Where do you see yourself in 10 years?'), I said that there will be four players. They will be Microsoft, Oracle, SAP and ... Google. I'm hoping to see myself with Google, the reason isn't obvious but I have mentioned it before:

Google does not charge the user. All of Google's software and hardware is at Google, behind locked doors, just like the IBM data center. To Google, you are a user, not an owner. You couldn't own the Google software because you would clog up Google's precious computers with your idiot code which couldn't possibly be as efficient and secure as Google's own code. They are masters of the algorithms and queue management. Google releases no product that doesn't scale to infinity. Google never gets a virus. Google is a free IBM datacenter. Google is the opposite of IBM despite all these similarities because its business model does not involve charging the user. Microsoft and IBM are the same because they charge the user. Microsoft makes you configure your machine and calls it freedom.

Furthermore, Google deals in human scale media. Google does not deliver compacted abstracted codes. Google delivers whole books, whole libraries of books. Google delivers satellite images of the whole planet. You are incapable of asking too much of Google, and their orientation is to deliver that all to you for free. Plus they take away all of the headaches that Gates delivered, the world of limitations of PCs. Clunky, insecure software written by anybody.

Google is one of the only computing products that hasn't bloated. Even Linux is running into controversy over kernel bloat. Pigs aren't flying, they're recompiling RPMs on their anti-Microsoft boxes. Same paradigm, same headaches. Free software approaches perfection, but there is no guarantee. Google guarantees, and that is why Google is rich and there are only two or three Linuxes left worth mentioning, all equally user poor.

But Google could charge the corporate user because Google has got what every enterprise desires: fundamentally unlimited scalability. They have the opportunity to license the way they do business and thereby revolutionize the IT business. Google is the ultimate ASP - they always have been, and they've been providing incredibly fast information for years now. Google can simply wait for every Tom, Dick and Ajax to write their little Web 2.0 app, and then determine whether or not it scales into a nice business model. They plug it into the Googleplex and voila.

It is not a stretch for me to see them doing this with 2 million spreadsheets, which is very likely what your corporation's finances are really running on despite MSFT's, Oracle's and SAP's protestations to the contrary.

So here's the question I posed to my BI noob interviewer. 'It takes about 15 minutes for you to find out whenever a car bomb goes off in Baghdad. How long does it take you to learn that your business has lost a customer?'. The answer to this question is generally sub-Google speed, which ultimately is unacceptable. The gap between real-time analytics is not so much technology as it is IT methodologies and captialization. This is something most companies would rather have, but not pay for up front. Some enterprising practitioners (like me) and some Google people could meet one day and come up with an ASP business model for enterprise BI. If and when that happens, MSFT, Oracle and SAP will find that they are a lot slower to the middle-tier webware market than they thought.

I am consistently surprised that people remain skeptical of fat clients. We are seeing 3GHz PCs with 100GB of disk and still people go Web. The hows and whys of that we can discuss later, but it is clearly a fait accompli. Whatever happened to port 80? It's now a thousand different ones for everyone's 'administrative client'. Client server has been swallowed up by very specific clients, ie browsers running interpreters. Whoda thunk? That being the case, applications development companies are essentially web development companies.

I have a hard time believing that people who built WebMethods or Vignette (nothing personal) are going to be developing the applications of the future. The narrow scopes of what IT organizations and (outsourced) consultants build these days unfortunately don't stand a chance against the big huge apps Google rolls out.

One final analogy. Once upon a time, Bloomberg relied on hardware. He had a near monopoly when it came to (bond trading I think). The smarts of that moved to PCs and the second revolution was exemplified in the day-trading bubble. (By the way, once upon a time the same was true of Metaphor DIS, specialty hardware to fat-client software to running on browsers in the BI space). Clearly the next evolution in tiered computing will be making browsers smart enough to do all fat clients do, cleaner. Now it's all about the scalability of the back-end. I have a hard time believing that the same decision makers who are gung-ho about punting to procurement departments when it comes to outsourcing staffing IT are going to think twice when offered the opportunity for 2 second response time without capital spending.

Microsoft Buys ProClarity: Smart Move

This is actually a fairly smart move by Microsoft as much as it is an admission that they don't understand the space or the customer base. Not that they care - now that they have a more or less complete front-end and lovely middle tier, their technology is end to end now.

ProClarity has smart people who understand MDX better than just about anybody in the industry. MDX is the multidimensional equivalent of SQL and it handles things like cross-tabs (to be non-technical about it) very nicely. In fact when I last checked a couple years ago (I'm certified, thank you) they were the only people in the industry who had built a competent MDX code generator. Their developers have scoped out a complete API for the product and I have little doubt that the whole thing will fit very nicely into the .NETframework. This means that .NET geeks will have no obstacles to building integrated analyic products into anything with an MSSQL back-end. This is something a small tech company with all the smarts cannot market by itself, but that Microsoft can definitely market.

The ProClarity suite of products has many more features than the average MSSQL customer uses and will enable people with lightweight ( 3 years) experience in the BI space to build relatively competant applications. So as an entre into your generic IT shop, they've got a headstart. But it also is plenty extensible. I wouldn't be surprised if a bulked up ProClarity group within MSFT didn't start building analytics into Great Plains under the Microsoft Dynamics label. It shouldn't take long for MS to integrate ProClarity into its product offering. All the hard stuff is already built. This is a slap in the face of Panorama Software, the Israeli company that helped usher in 'Plato' some of the original technology behind MSAS - MS' multidimensional server now embedded with MSSQL. It's also something of a slap to the interal people behind Microsoft Reporting Services, which although not very good for high-end apps is a good me-too against Business Objects, Cognos and Hyperion on the low end. It's reasonable to say, generically, that anybody who wants to run static format reports is a low-end analytical application anyway. Nobody expects much more who buys MSSQL anyway.

What's most interesting to me is whether or not ProClarity will continue to support MDX for Hyperion Essbase, the industry standard server for MDX on Linux & Unix boxes. Its got Yukon beat for scalability and does fabulous things on 64bit hardware. But as usual, the difficulty is getting people who understand high end BI to build with the pure technologies when so much of analytics is marketed in packages aimed at BPM and CRM etc. The last time I checked the ProClarity folks were very impressed with Essbase' Java API and how smoothly they were able to build. IE it worked the first time and did what it was supposed to do. The new MDX book out covers the differences between Essbase and MSAS, let's see how closely to the standards MSFT will remain. Microsoft really had to do this since Hyperion has already released their System 9 integrated platform 6 months ago. Oracle just recently made their announcement within the past week, and I suspect that MSFT is announcing their strategy to make this another tentacle in their .NET multiverse. This puts a great deal of pressure on Microstrategy, Business Objects and Cognos who are selling middle tiers and front-ends but don't own their own database technologies, but you never know, SAP might try something and Informatica is something of a wild card in this. IBM? Well they have the technology but who knows if they really care?

All in all this is a good way to seed future Yukon buyers with stuff that Microstrategy used to get away with. Since all the Yukon buyers are going to wait anyway, piling on some of the ProClarity stuff won't hurt MS.