Inktomi and Fast Search: Two Troubled Search Companies, One Lesson

May 8, 2012

I found the write up by Diego Basch interesting and thought provoking. I have a little experience with Inktomi. For the original system, the US government used Inktomi for the public facing index of US government unclassified information. ( is now

Inktomi had in 2000 a “ready to go” index of content from Dot Gov Web sites. The firm’s business model matched the needs of the US government. There were the normal contracting and technical hurdles for a modestly sized US government project with a fairly tight timeline. No big deal. Job done. Inktomi worked.

When I read “A Relevant Tale: How Google Killed Inktomi,” I thought the write up had some useful information. However, I don’t think Google killed Inktomi or any other search system. Google did not kill Fast Search & Transfer, Excite, HotBot, or any other search system in its rise to its alleged 65 percent share of the search market. (Google share is actually much higher, based on my analyses.)

excite_splash1996 copy

Excite’s early 1997 attempt at portalization. Can you spot the search box? Does this look like the current version of Google? Say, “No.” Now log into Google and run a query for rental car. Now do you see the similarity between the early portal craziness and the modern Google? I do.

What killed off these outfits was their business models. Let me explain using Inktomi and Fast Search as examples. I could cite other cases, but these two are okay for a free blog post for the two or three readers I have.

Inktomi, for whatever reason, concluded that people wanted to offer search, not do the heavy lifting. In the portal fever that was raging from 1998 to 2001, Web sites wanted to be the “front page” of the Internet. The result was that America Online, Excite, Lycos, and Yahoo among others jammed links on the splash page. At one time, I counted more than 60 links on the Excite home page. Once I hit 50 links, I quit counting. My eyes and patience can cope with three to five things. More than that, and I move on.

Inktomi’s analysts did the spreadsheet fever thing, making assumptions about how many Web sites would license Inktomi results, pay Inktomi’s fees, and generate revenue from the front page of the Internet craziness. The reality was that Inktomi did not have enough customers to support the cost of the spidering, bandwidth, investment in performance, research and development for precision and recall, and the other costs that are underestimated or just ignored. The result was the collapse of the company.

So what killed Inktomi was a failure of its business model. I am not into the woulda-coulda-shoulda school of thought. The fact is that the wrong business model doomed Inktomi. Assumptions about the number of customers and the amount those customers would pay were off. With revenue squeezed, cost cutting ensued. Value evaporated. Goner. Those Sun servers were expensive to buy and costly to scale. When more capacity was needed, Sun would often delay shipments for a month or more. When capacity is needed, it is needed now, not in a month or two. Bad business models reveal their weaknesses when costs cannot be controlled.

Now what about Fast Search & Transfer. That company had a snappy, useful search system. The demo site and then a potential alternative for Google was known as I never liked the name, but I did like the system. When 9/11 took place, had news available and accessible. Google, as I recall, did not. Score one for the Norwegian wizards.

In 2003, Fast Search & Transfer’s management made the decision to abandon the Web indexing and online advertising sector. The company sold Web property and search technology to Overture. Yahoo bought Overture and had the Inktomi system and the Fast Search system. Yahoo was clueless in the period from 2003 to 2006, and after settling the legal hassle with Google, Yahoo was unable to escape from portal craziness.

But what about Fast Search. The idea behind the sale of was to get cash so Fast Search & Transfer could build the killer enterprise search system. With the $140 million from the deal, Fast Search began to license its search system for enterprise use. The company made sales to some high profile outfits. However, the assumption that enterprises would license Fast Search and then go happily forward was incorrect. Organizations wanted to license Fast Search, have Fast Search engineers set it up and make it run quickly, and then have the search system do the marvelous functions the Fast Search sales professionals described. By mid 2005, there were customers who were grousing. By 2006, some customers who had signed contracts were demanding technical support and threatening to withhold payment. By 2007, some large licensees stopped paying. After some fancy footwork, Microsoft bought Fast Search & Transfer for $1.23 billion in 2008. Shortly after the deal, information became available revealing that Fast Search’s revenues were short of the figures reported to various regulatory outfits. Fast Search has become a component in Microsoft’s arsenal and investigation are still underway into various aspects of Fast Search’s financial reporting.

The failure of Fast Search & Transfer was due to a business model.

Let’s compare Inktomi and Fast Search. Both companies missed the importance of traffic to Web site owners. Web site owners wanted traffic and a search site which delivered traffic got love. If a Web site owner had to pay for traffic, Web site owners often did not care. Inktomi’s business model was blind to this revenue stream. Chasing a Web site owner to license search results just did not work. The Fast Search business model was based on the idea that the Fast Search system could be licensed to a company for a great deal of money. It could and it was, but the business model did not cope with a shortage of technical talent and the babysitting that many enterprise customers required. Fast Search sold off its Web site search to base a search business on a business model which just did not work for Fast Search.

Second, both companies collapsed rapidly. The reason each collapsed has to do with customers who would not play ball the Inktomi or Fast Search way. Without anyone in the game, both companies faced financial Armageddon. The end arrived quickly for both companies. The senior management of these companies struggled to solve problems but were unable to realize that the issue had more to do with the business model than delivering key word search. I heard the phrase “business blind spot” several years ago. Inktomi and Fast Search had a business blind spot when it came to solving the right problem. (I prefer William James’s phrase “a certain blindness.” Same idea, less zippy for sure.)

Third, both companies found ways to blame Google in part for implosion. In the case of Inktomi, I believe the article by Mr. Basch positions Google as the bad guy. I don’t buy it. I don’t think Google’s search was much better than Inktomi’s after relevance fixes were deployed. Inktomi had a business model in which its brand was “hidden” behind its licensees. Google developed brand traction; Inktomi did not. Google moved to paid search; Inktomi was marginalized. The the herd mentality took over. Most online searchers don’t know the right result from a lousy result. Google offered good enough search at a time when other companies were chasing portals and ignoring the importance of timely updates and snappy response.

Fast Search looked at the market and concluded enterprise search was a slam dunk. It was not and it is not. Consumer search is mass market and users don’t know if information is in the index, fresh, or accurate. Enetrprise search is highly specialized and some users know right away if a document is not in the index or findable.

Years ago, maybe 2007 or so, I heard a couple of former Fast Search executives explain that Google and its flawed Google Search Appliance was “freezing the market for a real search system.” Give me a break. Fast Search was horribly complex with crazy dependencies which meant a change in one script would cause a problem elsewhere in the system. Say what one will about the GSA in the period from 2002 to 2006, it would handle certain enterprise search and retrieval tasks without much hassle. Scaling and failure—those were different problems. But the basics were okay with the GSAs

Relevance in the enterprise means information which does not hurt revenues or increase risk. Fast Search’s executives did not understand that building a large enterprise business required Autonomy-type methods. Autonomy netted $10 billion; Fast Search is fading or has faded.

Business models. Sure, technology was a key factor. But the problems of these two companies boiled down to figuring out how to sell, make a profit, and grow. Neither did. Implosion in my book.

To sum up, a great deal has been made about Google’s brilliance. The search system was pretty good, but it arrived at a time when the search players were struggling for revenue and betting on the portal as the gold mine.

When Google’s management dropped the “we want to index for making” baloney, built an Overture type online advertising system, and let people buy traffic—Google faced essentiallyzero competition. Not now. Today Google has major league competition and it is not clear that Google’s management can deal with Amazon, Apple, Facebook, and Microsoft, among others. Solving a math problem and dealing with Android fragmentation, lawsuits on most continents, and softening online advertising performance require some different skills.

So Google’s apparent victory has more to do with its business model than its technology. Sure, Google depends on technology but so does Facebook and Amazon. The companies have business models enabled by the technology. Now the business model is preeminent in Google’s stiffest competitors. The “good enough” approach of Google is, for me, a key differentiator.

To wrap up, search technology is complex and has a long history. What makes a winner is the business model. Is Google the slayer of Web search competitors? Nah, Inktomi and Fast Search, among others  committed suicide by business model.

Stephen E Arnold, May 7, 2012

Sponsored by IKANOW


Comments are closed.

  • Archives

  • Recent Posts

  • Meta