• Home
• About
• Writers
• Landscape
• Wizards Index
• Fraud
• Wanna Be Like Larry?

More Allegations about Fast Search Impropriety

With legions of Microsoft Certified Resellers singing the praises of the FS4SP (formerly the Fast Search & Transfer search and retrieval system), sour notes are not easily heard. I don’t think many users of FS4SP know or care about the history of the company, its university-infused technology, or the machinations of the company’s senior management and Board of Directors. Ancient history.

I learned quite a bit in my close encounters with the Fast ESP technology. No, ESP does not mean extra sensory perception. ESP allegedly meant the enterprise search platform. Fast Search, before its purchase by Microsoft, was a platform, not a search engine. The idea was that the collection of components would be used to build applications in which search was an enabler. The idea was a good one, but search based applications required more than a PowerPoint to become a reality. The 64 bit Exalead system, developed long before Dassault acquired Exalead, was one of the first next generation, post Google systems to have a shot at delivering a viable search based application. (The race for SBAs, in my opinion, is not yet over, and there are some search vendors like PolySpot which are pushing in interesting new directions.) Fast Search was using marketing to pump up license deals. In fact, the marketing arm was more athletic than the firm’s engineering units. That, in my view, was the “issue” with Fast Search. Talk and demos were good. Implementation was a different platter of herring five ways.

Fast Search block diagram circa 2005. The system shows semantic and ontological components, asserts information on demand, and content publishing functions—all in addition to search and retrieval. Similar systems are marketed today, but hybrid content manipulation systems are often a work in progress in 2012. © Fast Search & Transfer

I once ended up with an interesting challenge resulting from a relatively large-scale, high-profile search implementation. Now you may have larger jobs than I typically get, but I was struggling with the shift from Inktomi to the AT&T Fast search system in order to index the public facing content of the US federal government.

Inktomi worked reasonably well, but the US government decided in its infinite wisdom to run a “free and open competition.” The usual suspects responded to the request for proposal and statement of work. I recall that “smarter than everyone else” Google ignored the US government’s requirements.

This image is from a presentation by Dr. Lervik about Digital Libraries, no date. The slide highlights the six key functions of the Fast Search search engine. These are extremely sophisticated functions. In 2012, only a few vendors can implement a single system with these operations running in the core platform. In fact, the wording could be used by search vendor marketers today. Fast Search knew where search was heading, but the future still has not arrived because writing about a function is different from delivering that function in a time and resource window which licensees can accommodate. © Fast Search & Transfer

Fast Search, with the guidance of savvy AT&T capture professionals, snagged the contract. That was a fateful procurement. Fast Search yielded to a team from Vivisimo and Microsoft. Then Microsoft bought Fast Search, and the US government began its shift to open source search. Another consequence is that Google, as you may know, never caught on in the US Federal government in the manner that I and others assumed the company would. I often wonder what would have happened if Google’s capture team had responded to the statement of work instead of pointing out that the requirements were not interesting.

Nevertheless, the Fast Search implementation was darned exciting, and I gathered some fascinating, first-hand information about index times, system performance, tuning, and customization. Ah, valuable lessons which are ignored by many 30 somethings today.

In the salad days of Fast Search, the Norwegian company abandoned Web search to Google and decided to focus on enterprise search. At that time, 2002 and 2003, Fast Search had the cash from its sale of its Web search technology to Overture, later Yahoo. The idea was that Fast Search was better, faster, cheaper, and in every possible way, a more appropriate system for the enterprise than Autonomy, Convera, Endeca, Oracle, and Verity. What interested me was that Fast Search had a variable alternative to Google and even possessed some advertising capability. A fateful decision it was to abandon Web search and advertising based revenue for the enterprise market.

Google figured out that running Web search and ads was like a SpanDex outfit—one size would fit everyone. The enterprise market was nothing more than unique search installations. The financial challenge of tailoring a solution to many different customer use cases remains the single biggest problem for enterprise search vendors. Even the verticalization in customer support or business intelligence requires expensive, time consuming tailoring. Fast Search ignored the marketing and engineering cost implications of selling one offs. Licensees then and now don’t know much about search costs. Fast Search reported quarter on quarter growth that struck me as surprisingly robust. Even AT&T believed the Fast Search marketing pitch. One AT&T project manager told me, “The Fast Search CTO is so wonderfully pleasant.” Yep, pleasant.

There I was in a post 9/11 project which was under scrutiny from the White House, various Government Accountability observers, and the General Services Administration. I ended up involved in both the procurement process which would morph into implementation and then after a few years and IV&V job on this project due to some interesting accounting maneuvers. (IV&V is government speak for investigations related to technology implementations. The acronym stands for Independent Verification and Validation. That work is not public even today.)

This is a diagram used by Dr. Lervik to show how Fast Search could parse a document and extract key elements of that document. The idea was that such parsing would permit text mining. Google’s Alon Halevy was working independently on a somewhat similar system. Dr. Halevy’s work has not yet become a user facing service. Google continues to invest in sophisticated content processing. Once again, Fast Search had a good idea, but the costs and complexity associated with certain ideas can stretch a vendor’s financial resources. The appetite for money in the search and retrieval business is significant. © Fast Search & Transfer

When the news of Microsoft’s purchase of Fast Search for an astounding at the time $1.2 billion, I asked myself, “Who in the heck at Microsoft did the due diligence on this deal.” I had picked up rumblings of dissatisfaction among Fast Search licensees after my brush with the system in the 2002 to 2004 period. As I recall, I heard about some technical issues on a job for a newspaper in London. Next, I provided input to a problem at a German publishing company. In short, I was asked to write about Fast.

I was giving a talk in London in October 2008 when I learned that the Norwegian authorities had visited the Fast Search offices in Oslo. The purpose of the authorities’ visit was to gather information about the company’s finances. At the time of the 2008 visit, Microsoft was in the driver’s seat and seemed happy to act as if there were zero problems with the company, its technology, or Fast Search license deals. Microsoft, in my opinion, has just moved on, marketing the heck out of the Fast Search technology.

Imagine my surprise when I learned from my contacts in Norway that the Web site DN.no (Business Day) ran this story: “Fast-topper tiltalt for regnskapsbrudd og markedsmanipulasjon.” You can read it in Norwegian, have a friendly Norwegian librarian translate it, or use the good-enough Google system. Here’s the key passage:

Norwegian authorities have accused former CEO John Markus Lervik and former sales director Torbjorn Kanestrøm company FAST Search & Transfer ASA (Fast) for violations of the Accounting and market manipulation.

The article digs into the allegations and includes some interesting historical references. The accused search wizard, John Lervik allegedly said:

I had nothing to gain by manipulating accounts. I had no stake in the company. I wanted a stake, and stopped because I did not. To find out what the problem with the accounts, you must look at who benefited from it. And it was not me, he told Business Today that time.

I don’t have any fungible information on this most recent matter.

My view remains:

• Fast Search was reporting robust revenues up to the time of the sale to Microsoft, then revenues were questioned and, as I recall, restated at a lower figure
• Customers were grousing that the 64 bit version of Fast Search was delayed
• The use of proprietary, open source, and partners’ code was adding to the time and cost of deployments
• Engineering resources were not sufficiently deep to respond to some licensees’ requests for prompt service.

This is a Dr. Lervik diagram summarizing how Fast Search in the period before the Microsoft acquisition could perform linguistic query analysis. Since the acquisition, the technology continues to require third party add ins and engineering services to deliver the Fast “vision.” © Fast Search & Transfer

The legacy of Fast, like the legacy of Google, is quite different from the information now circulating about the system. As I write this, Microsoft’s integrators are finding that consulting and engineering services, components to make SharePoint search hum, and off-the-shelf snap ins are a large, booming business. Microsoft has, I have heard, more than 100 million SharePoint licensees. Each of these is a gold mine for those who can make SharePoint search work like Autonomy or any other established search system.

I may write about the legacy of Fast Search, but for now, recalling the rise, sale, and possible legal hassle about Fast Search is a reminder to those who think search is simple. Search is tough. Generating revenue in the enterprise market is tougher. As enterprise search is commoditized, one must be aware of what costs accompany any search and retrieval implementation. Some costs are technical. Some are informational. Others are personal. Search can bite and bite hard.

Stephen E Arnold, March 8, 2012

Sponsored by Pandia.com

Comments

Copyright © 2008 Beyond Search · ArnoldIT.com · Disclaimer · Revolution News theme by Brian Gardner · Log in