Enterprise Search: Hidden and Intentional Limitations
November 29, 2014
Several years ago, an ArnoldIT team tackled a content management search system that had three characteristics: [1] The system could not locate content a writer saved to the system. [2] The search results lacked relevance that made sense to the user; that is, time sequences for versions of the article were neither in ascending or descending order. [3] Response time was sluggish. Let’s look at each of these obvious problems.
The user wanted to access a document that required final touch ups. The article was not in a result that even when the writer entered the full title of the article. Examination of the system revealed that it lacked keyword search, relying on a training set of documents that did not contain some of the words in the title. The fix was to teach the user to search for articles using words and concepts that appeared in the body of the article. We also identified an indexing latency problem. The system lacked sufficient resources so recent material was not in an index until the indexing system caught up. The organization did not have the money to add additional resources, so the writers were told, “Live with it.”
What is that vendor doing amidst the prospects? A happy quack to http://bit.ly/1CrUc81
The chaos in the relevance function was a result of a configuration error during an upgrade and subsequent reindexing. The fix was to reconfigure the system and reindex the content. Keep in mind that indexing required more than two weeks. The attitude of the client was, “It is what it is.”
The third problem was the response time. This particular system used a product from a Canadian vendor of search and content management systems. The firm acquires companies and then “milks” the system. The idea is that updating and bug fixing are expensive. The problem was exacerbated because the search system was a stub provided by another vendor. The result was that in order to get more robust performance, the client had to upgrade the OEM search system AND add computers, memory, and network infrastructure. The client lacked the money to take these actions.
What were the hidden limitations of this enterprise search system?
On the client side, there was a lack of understanding of the interdependencies of a complex system. The client lacked expertise and, more importantly, money to address the problems of an “as is” infrastructure. Moving the operation to the cloud was not possible due to security concerns, knowledge of what to do and how to do it, and initiative on the management team. The conclusion was, “The system is good enough.”
On the vendor side, the marketing team closing the deal did not disclose that the OEM search system was a stub designed to upsell the licensee the seven figure “fix”. The vendor also did not reveal that funds were not being invested in the system due to the vendor’s “milk the cow” philosophy.
I point out this situation because it applies to many vendors of enterprise search systems. The object of the game on the client side is to get a good enough system at the best possible price. Senior managers are rarely experts in search and often search Google anyway.
The vendor has zero incentive to change its business practices. Even with low cost options available, once a prospect becomes a customer, lock in usually keeps the account alive. When a switch is in the wind, the search vendor plays the “legacy” card, pointing out that there are some people who need the older system. As a result, the licensing organization ends up with multiple search systems. The demand for money just goes up and findability remains a bit of a challenge for employees.
I do not see a fix under the present enterprise search business model. Education does not work. Search conferences dodge the tough issues, preferring to pander to vendors who buy exhibit stands and sponsor lunch.
Something different is needed: Different vendors, different approaches, and different assemblies of technology.
That’s why next generation information access is going to pose a threat to companies that pitch enterprise search disguised as customer support, business intelligence, analysis systems, and eDiscovery tools.
At some point, the NGIA vendors will emerge as the go-to vendors. Like the promising but now quiet outfits like Hakia and Autonomy, search as it has been practiced for decades is rapidly becoming a digital Antikythera mechanism.
Stephen E Arnold, November 29, 2014
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
