Enterprise Search and Train Wrecks

May 7, 2008

After I completed my interview with the Intelligenx executives, I thought about one of their comments. Iqbal Talib said, “We have many clients who want a point solution, not an enterprise solution”. An executive at Avalon Consulting wrote me today and echoed the Intelligenx comment.

Enterprise search may be a train wreck for more than half of the people who use today’s most popular systems. The Big Name vendors can grouse, stomp, and sneer at this assertion. Reality: Most of these systems disappoint their licensees. When a search system “goes off the rails”, the consequences can be unexpected.


When an enterprise system goes off the rails, the damage is considerable. Even worse, moving the wreckage out of the way is real work. But even more difficult is earning back the confidence of the passengers.

A Case Example

A major European news organization licensed a Big Name system. The company ponied up a down payment and asked for a fast-cycle installation. After six months of dithering, the Big Name admitted that it did not have an engineer available who could perform the installation and customization the paying customer wanted.

The news organization pulled the plug. The company then licensed one of the up-and-coming systems profiled in Beyond Search. The revamped system was available in less than three weeks at a fraction of the cost for the Big Name system.

The new system works, and it has become a showcase for the news organization. For the Big Name, the loss of the account eroded already shaky finances and became the talk of cocktail parties at industry functions.

Ever wonder how much churn Big Name enterprise search vendors experience in a year? You can get a good idea by comparing the customer lists of the best-known enterprise search vendors. The overlap is remarkable because large companies work their way through the systems. Now more are turning to up-and-coming vendors’ systems. The Big Names are facing some sales push back. Take a look at the financials for publicly traded search vendors. Look for days-sales-outstanding data. Look at the cash reserves. Look at the footnotes about restating financials.

What you may find is that fancy dancing is endemic.

How Many Search Systems Does One Company Need?

What haunts me is the overlap among vendors. Early in 2003, I conducted a poll of Fortune 1000 companies. The methodology was simple: I sent an email with several basic questions to people whom I knew at 150 different large organizations. I received a response rate of about 70 percent, which was remarkable. One question I asked five years ago was, “How many enterprise search systems do you have?”

The answer was five or more.

In my research for the first three editions of the Enterprise Search Report and my new Beyond Search study, I kept a list of the reasons for the proliferation of “enterprise search” systems.

The obvious conclusion to me was that “enterprise search systems” did not do enterprise search. Search systems tackled specific information access problems. What organizations wanted were point solutions that worked to solve a problem. If a system could handle another challenge or two without too much hassle, that was a bonus.

But the data are clear. Instead of trying to warp a search system to fit another information challenge, most organizations found it simpler to license yet another enterprise search system. So competitors like Autonomy, Convera (which has gone gently into that good night of enterprise search and morphed into a “vertical search company” whatever that means), Endeca, Fast Search & Transfer, IBM (the company with many different ways to “do” search), Microsoft, Oracle, and SAP bumped into one another frequently. It also explains the frequent statements of search vendors to me, “We beat those guys at Company X.” Everyone was beating everyone else because large companies seemed to be ready to buy a half dozen or more search solutions.

Why Appliances and Hosting Are Getting More Traction

In my run up to Beyond Search, I was quite interested in companies selling appliances. These are ready-to-run gizmos that a licensee plugs in, configures, and makes available as a search toaster. Google and Thunderstone sell search appliance, but you can find products on offer from Exegy, Index Engines, Planet Technologies. (Note: Exegy is profiled in Beyond Search because of its novel approach to fast content processing, not its marketing savvy.)

The proof is the success of the appliances from Google and Thunderstone. The strong interest in Lucene and other open source search systems. Blossom Software, a search vendor with a hosted solution, is having a banner year. In my opinion, the SAS Institute’s acquisition of Teragram underscores the opportunity in content processing. SAS has built a good business with its core analytics business. The company would not be building out its text capabilities unless its management had data indicating that there were clients who wanted something beyond search. In short, SAS’s purchasing of Teragram underscores the problems traditional search vendors have created for their clients and themselves. If incumbents were doing the job, Google wouldn’t have had such an easy time of making more than 8,500 sales. Blossom Software wouldn’t be having a banner year. And up and coming vendors wouldn’t be forcing Microsoft to pony up $1.2 billion to get back into the search game.

Causes of Failure

In the case about the major European news company, the cause of failure was simple. The search system vendor didn’t have the staff on hand to do the niggly configuration and tuning its technology demands.

There are other causes of failure, and you probably know them better than I do. When I write essays for this Web log, I’m able to pick up lists of reasons why search disappoints from eight or nine years ago. The items almost always apply to today’s implementations. For example, licensees don’t understand how complicated search and text processing is. I don’t think in house information technology professionals understand this issue until they have been bloodied in battle. The vendors deliver a “work in progress”, not a product. Even worse, there’s no “product”; there’s a list of options, and these have to be assembled like a physics experiemtn in graduate school. The company doesn’t understand the meaning of “appropriate resources”. When a system slows down, money and hardware and programming and work flow fixes are required. None of these needs can be met over night. Time equals money. Management gets impatient and buys a Google Search Appliance. So, Google’s 8,500 sales are pointers to 8,500 organizations who found traditional enterprise search unsatisfactory to some degree. In case you had not noticed, Google is not very good at sales. I can’t imagine how many search appliances the company would sell if it returned phone calls and showed up at meetings with someone who could talk to customers in a regular way, not a Googley way.

Where We Are Now

PR alert. I’m not going to name your company, so don’t send me disingenuous emails wanting to brief me on your employer’s system. Pander elsewhere, please.

When enterprise search systems crash and burn, the damage is significant. Costs sky rocket. Even worse, employees have to scramble to find the information they need to do their work. If search doesn’t deliver, an organization runs the risk of making uninformed decisions. In today’s world, the margin for error is decreasing rapidly. Bad search doesn’t just cost money. Bad search can hobble an organization. Predators target the lame.

So, here’s where a great many organizations stand:

  1. The enterprise search system is, at best, okay. Point solutions from vendors who deliver a solution that works are making life tough for the Big Name vendors. Gentle reader, there’s a reason Google has more than 8,500 search appliance customers. The GOOG’s search toaster works well enough.
  2. The all-encompassing Big Name system requires an organization to change its procedures. Users want the systems to adapt to their needs. Thus, there is a fundamental mismatch.
  3. Information needs in an organization are dynamic. Big Name systems require time-consuming coddling. Speed wins. Little wonder that hot, up and coming system vendors like Coveo, ISYS Search Software, Siderean Software, and a handful of others are experiencing strong sales.

And what about the vendors?

Everytime I see a meaningless news release announcing a deal with an unknown sports club with a Web site in a country unknown to geographers in Harrod’s Creek, I think, “These guys are really struggling.”

When I read about a “new” way to license multi media search, I conclude, “Okay, that outfit is doing whatever it can to generate some cash.”

When I get a call from an expert who wants my comment on a whizzy new feature that will revolutionize search, I conclude, “Not only is the caller uninformed, the company shoveling this baloney should be selling bankruptcy advice to people on the fast track to debtors’ prison.”

Now What?

In enterprise search, organizations are showing keen interest in the up-and-coming vendors’ identified in my study Beyond Search.

In my opinion, the strong growth reported by the up-and-coming vendors is bad news for some of the Big Name vendors. I think the Big Name vendors will take these actions more frequently and with greater aggressiveness:

  1. PR releases about meaningless deals
  2. Speeches about “break throughs” that won’t address the core problems of complexity but increase the chances that moving parts in the system will jam up and break
  3. Product and service announcements that are “me too” statements. Everyone federates. Everyone analyzed the “deep Web.” Everyone extracts entities and classifies. Wrong. There functions must work in a real world environment not on a demo that fakes out the potential customer.
  4. Repositioning statements. Search means information access. A company that offers a utility for importing content is not an enterprise vendor. Conversely, an enterprise search vendor sells like a street vendor. Anything that can be piled on the cart is the product. Writing around what the software does is not helpful to anyone in the long run.

If your organization is saddled with a Big Name search system that doesn’t work very well, you are looking for a way out. My concern is for organizations that don’t know their Big Name system doesn’t deliver. I call this a “cloud of unknowing”. Vendors work hard to keep the licensee in a fog.

Here’s what I heard one vendor tell a person standing at a trade show booth last week:

Oh, that issue. Fixes are imminent. The new connector will resolve a bottleneck. Our recommendation is to upgrade the search server to a quad core. That easy tweak will increase throughput dramatically. The next release will contain a greatly improved point-and-click interface.

Now what means looking for a better, faster, cheaper solution. The idea of forcing the licensee to pick two of these attributes won’t work in today’s constrained business environment.


The significant development is that Microsoft bought Fast Search. Yesterday (May 6, 2008), I spoke with a New York journalist who was trying to work out if the $1.2 billion price tag was fair. In my opinion, it was a great payday for Fast Search & Transfer’s Board of Directors and some of the management team. Microsoft, in my view, has only a partial understanding of what Fast ESP (enterprise search platform) does well and how it “fits” into an organization. In my opinion, I think Microsoft’s big deal mentality contributed to a modest over valuation of Fast Search. My back of pencil calculation which accounts for some financial adjustments suggests a price tag of 15 to 20 times actual revenues in 2007. The official figure is closer to 10 times 2007’s reported revenues. But, what’s an order of magnitude when you’re doing New Zealand sized deals?

Forget the $1.2 billion already.

The purchase of Fast Search by Microsoft makes it clear that SharePoint’s “native” search can’t handle certain types of SharePoint search tasks. With “appropriate resources” Fast ESP can handle big indexing jobs. At this moment in time, Fast Search’s John Lervik has an opportunity to

  • Keep certain SharePoint customers from wandering off the reservation by licensing systems from Coveo, Exalead, ISYS Search Software, and Siderean Software, among others
  • Craft a cohesive search platform at Microsoft. My odds on this happening are 60 to one. Cohesive anything at Microsoft is a long shot. The interface across Microsoft Office varies from applicaton to application. Search is trickier than an interface
  • Prove that Fast ESP is a Google-killer. Work on this front needs to move with alacrity. The GOOG isn’t taking a holiday so Microsoft can catch up.

But it’s not just Microsoft that’s having to change.

The cash infusion from Intel and SAP into Endeca gives that company an opportunity to move from its $100 million base in 2007.

Autonomy, on the strength of its most recent quarter, should be able to blow by Google and regain the coveted crown of Top Enterprise Search Dog. PR mavens and Enron-inspired analysts take note: I don’t care that Google reported $188 million in enterprise revenues. My sources tell me that I have to figure in the “other” money from maps, deals with institutions, partner payments, and similar sources of enterprise revenue. When I do these math tricks, Google pops out as the largest vendor of enterprise search. Feel free to agree or disagree. I’m sticking by my analysis.

Meanwhile, the up and coming vendors will exert pressure on the Big Names, including IBM and Oracle. Adding spice to the mix are companies with solid revenues encroaching on Big Names and up and comers. SAS Institute and SPSS are, for all practical purposes, selling enterprise search solutions.

What’s obvious is that the enterprise search sector is caught in a seismic shift. News of train wrecks creates opportunities for vendors who arrive with no history of catastrophic failure. Those with alternate means of transportation can work around the twisted rails and scattered chunks of metal.

Enjoy the train ride. I will walk, thank you.

Stephen Arnold, May 7, 2008


3 Responses to “Enterprise Search and Train Wrecks”

  1. Advice on business on May 8th, 2008 12:23 am

    Your posts are always informative and helpful. Thanks for all your hard work on this one!

  2. Stephen E. Arnold on May 8th, 2008 7:54 am

    I appreciate the kind words. Keep in mind the information in the Beyond Search Web log is from my earlier presentations and articles. The new, hot stuff appears in my for-fee studies like Google Version 2.0 (which provides a glimpse of what Google’s engineers are developing for next-generation services) and Beyond Search, which provides tips for remediating a screwed up enterprise search system plus profiles with evaluations of 24 vendors challenging Autonomy, Endeca, Fast Search & Transfer, Oracle, and SAP among others in the text retrieval market. Don’t be hesitant to take issue with my opinions. I’m open to different viewpoints, and I’m not yet too old to learn. I just don’t learn quickly. Old goose and all that.
    Stephen Arnold, May 8, 2008, 9 am Eastern time

  3. Fast Cash, Faster Crash : Beyond Search on September 15th, 2011 3:48 pm

    […] Enterprise Search and Train Wrecks : Beyond Search http://arnoldit.com/wordpress/2008/05/07/enterprise-search-and-train-wrecks/ […]

  • Archives

  • Recent Posts

  • Meta