Enterprise Search: Finding Flounders Floundering
September 9, 2013
A founder is a flat fish. In today’s whiz kid world, “flounder” does not make one drool for a fish stew. “Flounder” means to the Free Dictionary, “to move or act clumsily and in confusion.” I made the connection to search as a result of a seemingly innocuous discussion on LinkedIn about improving search.
A flounder not yet out of water.
I am not sure about the rules for linking to LinkedIn content. I have to watch my Ps and Qs because two of the goslings and I learned on Friday, September 6, 2013, that some of the queries I launched from my research computer were not processed by Slideshare. Was this a glitch or some intentional action? I don’t know. To be on the safe side, I will not link to the thread called “How to manage queries having no relevant answers but still matching some terms.” If you are a LinkedIn customer, you can log in and locate the discussion using the LinkedIn “finding” system. How well with that work out for you? Well, that’s another search topic.
To recap the thread, a LinkedIn customer is responsible for an Intranet search system. When its users run a query, the system produces a results list which do not answer some users’ questions. There is term matching, but the content is not on point. I no longer like to beat the drum for precision and recall. We are now in the era of good enough search. Few take the time to create a vetted content inventory. When the search system is rolled out, no one really knows what’s “in” the index. The point that a query contains terms which match some content but makes users grouse is not new.
Caught by an unhappy user who happens to be the CFO figuring out why so much money was spent for a search system that did not work.
The fix, of course, is like trying to refuel an old fashioned propeller driven aircraft with a somewhat more modern jet powered tanker. The job is going to be tricky and may end with some excitement. Jets and prop driven aircraft like enterprise search and quick mixes might not be a happy combination like peanut butter and jelly.
In 2004, then Googler Dave Girouard said in eCommerce Times:
“The funny part is it’s easier to find box scores from the 1957 World Series than it is to find last quarter’s sales presentation in the enterprise. While Web search has gotten really good, enterprise search has stagnated, and that’s why we really believe it’s a problem that needs to be solved and that Google has a unique set of capabilities to solve it.”
Well, Mr. Girouard has moved on and Google is advertising on LinkedIn for yet another wizard to work on enterprise search. If Google cannot knock the ball out of the park, who can? Is HP Autonomy the go-to system? What about a low-cost option like dtSearch? Why not download Elasticsearch, Constellio, of one of the other open source solutions? Maybe a company should embrace a predictive solution from Agilex or Palantir?
The fact is that some basics have to be nailed down. Unfortunately the “basics” are often time consuming and difficult to work out. Who has time today? Most of the folks in the enterprise search game are in a big hurry or don’t have the expertise to do the “basics.” Not surprisingly, the LinkedIn type of discussions focus on quick fixes. I interpret these posts as “I have to do something to save my job.”
Here’s what I posted on LinkedIn this morning. Feel free to attack my comment, not me, please, using LinkedIn or the comments section of this blog:
Research we conducted last year revealed that most major “findability” systems are increasingly alike. The differentiators are connectors, shortcuts to speed certain processes, and interfaces. The relevance and precision scores are in line with TREC data. Further, user dissatisfaction with search, regardless of system, ranges from 50 percent to as much as 65 percent. After 40 or 50 years of flailing in enterprise search, there are several questions to address, based on my team’s work: (1) What is the specific search problem that must be solved? Chemical structure search is different from AutoCAD document search and so on across structured and unstructured information.) (2) What is the content available within the organization to meet the user’s need? If the content is not available, what steps can be taken to provide the content? (3) What is the user’s requirement for timeliness, comprehensiveness, outputs, etc.? (4) What is the budget for the system? (5) Does management support the solution? (6) What systems match the quite narrowly defined requirements? (7) Which systems are candidates given the resource constraints and team capabilities? There are other questions to address. When a search system runs into trouble, “tactics” won’t deal with the problem. At some point, the accounting types will call a time out, push the reset button, get a new team and try again. Has this process delivered user satisfaction? Perhaps more than “tactics” are warranted? Stephen E Arnold, September 9, 2013
First, everyone is an expert in search. The reality is that knowledge about what is required to answer a user’s question is not particularly deep in my experience. The notion of a content inventory and content acquisition * before * rolling out a search engine is viewed as stupid as Marx Brothers jokes told at Pauley Shore’s comedy club in LA last night.
Second, the systems are increasingly alike. Therefore, the problems which talk about unhappy users are, in my opinion, not solvable by changing from System A to System B. The effort required is a distraction. Why not put that effort into the basics? The answer, we have learned, is that the basics are not needed in the view of Google experts and those who chatter about fancy math “predicting” or “discovering” what’s important. Well, if the content does not contain the answer or is not accurate, how valid are those outputs, folks?
Third, most of the experts in search are not sure what to do. The high profile outfits recommend systems but do not install them. The “integrators” know how to address problems and dance away from telling the Board of Directors that the company has blown big money and risks incorrect business decisions on existing systems. Who wants to risk losing a client? Just run up the billings and move on, right?
My personal view of search is that search remains a difficult problem. As I will point out in my lecture at the ISS conference, finding information today is more difficult than it was 20 years ago. Talk about an unpopular view: I can see the hands flapping during Q&A now.
Try this simple test. Find in your organization an email sent 10 minutes ago about the CFO’s presentation at the upcoming quarterly meeting to discuss sales. Let me know why this type of information is not available to you? My hunch is that the existing search system filters content and lacks the capability of updating its indexes within 10 minutes. In short, existing search systems have some blind spots.
As William James, the ever popular brother of that exciting writer Henry James, said:
The art of being wise is the art of knowing what to overlook.
Would William James overlook the basics? I think not. But if one does not know what to overlook, lousy search systems result. Tactics cannot fix missteps which are fundamental.
Stephen E Arnold, September 9, 2013