No Wonder Search Is Broken. Software Does Not Work.
September 17, 2012
Several years ago, I ran across a Microsoft centric podcast hosted by an affable American, Scott Hanselman. At the time he worked for a company developing software for the enterprise. Then I think he started working at Microsoft and I lost track of him.
I read “Everything’s Broken and Nobody’s Upset.” The author was Scott Hanselman, who is “a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee.”
The article is a list of bullet points. Each bullet point identifies a range of software problems. Some of these were familiar; for example, iPhoto’s choking on large numbers of pictures on my wife’s new Mac laptop. Others were unknown to me; for example, the lousy performance of Gmail. Hopefully Eric Brewer, founder of Inktomi, can help improve the performance of some Google services.
Answer to the Google query “Why are Americans…”
The problems, Mr. Hanselman, identifies can be fixed. He writes:
Here we are in 2012 in a world of open standards on an open network, with angle brackets and curly braces flying at gigabit speeds and it’s all a mess. Everyone sucks, equally and completely.
- Is this a speed problem? Are we feeling we have to develop too fast and loose?
- Is it a quality issue? Have we forgotten the art and science of Software QA?
- Is it a people problem? Are folks just not passionate about their software enough to fix it?
I think it’s all of the above. We need to care and we need the collective will to fix it.
My reaction was surprise. I know search, content processing, and Fancy Dan analytics do not work as advertised, as expected, or, in some cases, very well despite the best efforts of rocket scientists.
The idea that the broad world of software is broken was an interesting idea. Last week, I struggled with a client who could not explain what its new technology actually delivered to a user. The reason was that the words the person was using did not match what the new software widget actually did. Maybe the rush to come up with clever marketing catchphrases is more important than solving a problem for a user?
In the three disciplines we monitor—search, content processing, and analytics—I do not have a broad method for remediating “broken” software. My team and I have found that the approach outlined by Martin White and I in Successful Enterprise Search Management is just ignored by those implementing search. I can’t speak for Martin, but my experience is that the people who want to implement a search, content processing or analytics system demonstrate these characteristics. These items are not universally shared, but I have gathered the most frequent actions and statements over the last year for the list. The reason for lousy search-related systems:
- Short cuts only, please. Another consultant explained that buying third party components was cheaper, quicker, and easier than looking at the existing search related system
- Something for nothing. The idea is that a free system is going to save the day.
- New is better. The perception that a new system from a different vendor would solve the findability problem because it was different
- We are too busy. The belief that talking to the users of a system was a waste of time. The typical statement about this can be summarized, “Users don’t know what they want or need.”
- No appetite for grunt work. This is an entitlement problem because figuring out metrics like content volume, processing issues for content normalization, and reviewing candidate term lists is not their job or too hard.
- No knowledge. This is a weird problem caused in part by point-and-click interfaces or predictive systems like Google’s. Those who should know about search related issues do not. Therefore, education is needed. Like recalcitrant 6th graders, the effort required to learn is not there.
- Looking for greener pastures. Many of those working on search related projects are looking to jump to a different and higher paying job in the organization or leave the company to do a start up. As a result, search related projects are irrelevant.
The problem in search, therefore, is not the technology. Most of the systems are essentially the same as those which have been available for decades. Yes, decades. Precision and recall remain in the 80 percent range. Predictive systems chop down data sets to more usable chunks but prediction is a hit and miss game. Automated indexing requires a human to keep the system on track.
The problem is anchored in humans: Their knowledge, their ability to prioritize search related tasks, their willingness to learn. Net net: Software is not getting much better, but it is prettier than a blinking dot on a VAX terminal. Better? Nah. Upset? Nope, there are distractions and Facebook pals to provide assurances that everything is A-OK.
Stephen E Arnold, September 17, 2012
Sponsored by Augmentext
Comments
One Response to “No Wonder Search Is Broken. Software Does Not Work.”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
Totally agree with you. Surveys of ERP implementations indicate a level of at a minimum one member of a support team for every 100 users. Even in search implementations across 50,000+ users there are rarely more than a couple of people looking after search issues.
Although the focus is on ‘big data’ reading the recent survey by Oracle on the costs of not being able to find information makes depressing reading.
http://www.oracle.com/us/industries/industry-scorecard-1683398.html
A well-planned and supported search implementation in all companies with more than 1000 people could transform the economy of any country.