Navigation bar  
Search AIT

Only match whole words

An Interview with Martin White

Martin White , Managing Director, of Intranet Focus Ltd. in the United Kingdom, is one of the leading experts on information access, information governance, and Intranets. Mr. White’s Web site is here. He and Stephen E. Arnold completed a study about what makes an enterprise search system successful. The study, published by Galatea in the United Kingdom, extracts the lessons learned from the decades of experience, dozens of projects, and numerous discussions with vendors, licensees, and colleagues worldwide. This interview is with Mr. White and presented as a conversation with him. Mr. Arnold and Mr. White collaborated on the responses and decided that for the sake of readability, to make the dialog reflect a conversation between Mr. Arnold and Mr. White. The interview was conducted on November 10, 2008.

You can learn more about the study at or click here.

Why is search in an organization important?

Usually without realizing it companies have invested huge amounts of staff time (and therefore money) in creating information but it is largely hidden away in multiple applications and shared drives. Although most of these have some form of folder/file structure these have long been overtaken by changes in business objectives. Search, when well implemented, offers at least some chance of providing the employee with the ability to find information without having to know where the information is stored or what it is called.

Users are annoyed, maybe dissatisfied, with search in their organization. Why? We have been doing key word search for 25 years, right?

Nearer 50 years if you go back to the Dialog and SDC timeshare search services of the late 1960’s. Keywords may help to find a few relevant references but in an enterprise situation recall (finding as many as possible of the relevant documents) is perhaps even more important. Enterprise searches are way too complex for just a keyword or two as you are dealing with millions of documents written by thousands of people with their own particular way of expressing concepts. It’s concepts like ‘new product development’ that keywords really start to fail.

What's the lowest risk way to deploy search that does not annoy users?

Do not ask them what they want the search engine to do, but find out what decisions they need to make and then work out what information they need. Finally work out what questions they are likely to ask to try to find the information they need to make the decisions. Note I said "information" and not "document" or "data base."

When search goes wrong, what are the warning signs? Costs? Lack of system use? How do successful organizations respond to this situation?

Lack of system use may not be immediately obvious. For a start around 15% of staff will be new each year and they don’t know the system is useless. A systematic approach to not only analyzing search logs but then fixing the problems is called for. People often don’t complain about poor search as they think the problem is their lack of skill in using the application.

How can an organization fix search when management is the real problem? A procurement team can be powerless, yet poised to take the blame. Is there a built in failure momentum?

When search fails it is almost always a result of the fact that the organization has no commitment to effective information management at both a strategic and operational level. This is now changing, as indicated by a recent Economist Intelligence Unit report, and this is good news for the search industry.

Is there a constructive role for vendors to play in a search procurement? What is it?

Vendors can do three things. First show best practice in search on their sites. Second they can realize that the people thinking of buying their products have probably never bought a search engine before. Third they can stop suggesting that search will be perfect, as that is dependent a lot of variables outside of their control, notably content quality.

You have written a new book about the problems associated with managing a successful search deployment? Why management? Why not technology?

Very rarely is poor search a result solely of poor technology. It is all about effective management of the entire search procurement, installation and implementation processes.

Search is very complex. Do you think most of those involved in enterprise search on the buyer side understand the complexities? How can this problem be addressed in an efficacious and timely way?

After 35 years in search I don’t understand all the technology. Only you do. The key is to test out the search application against clear requirements and using well designed sets of test documents. So long as someone on the procurement and implementation teams knows how search works that will suffice, but this is where consultants can play an important role.

Someone told me, "Business intelligence is the new search." Do you agree? What is the "new search" in your opinion?

Look at all those financial institutions with their BI applications. Did it stop them making a fool of themselves and us over sub-prime loans. BI is only as good as the way in which the correlations are set up and usually that is poorly. To me the new search is when search is all but invisible – embedded in a workflow process.

How does a resource constrained organized deal with the problem of controlling costs for such unknowns as next month's network demand, server capacity, query latency, and content processing bottlenecks? Talk and puffery don't resolve fungible technical issues. What does when there is a shortage of resources?

For some years capacity planners have played a key role in ensuring that systems do not implode. They have the skills to develop models for search applications but because CTOs have no idea of the variables and risks of dysfunctional search no new models are being developed.

Has enterprise search become to complex for all but the largest, wealthiest organizations?

The largest and wealthiest organizations are no more likely to make a success of enterprise search than smaller organizations. Enterprise search is no more complex than a large ERP application, but ERP applications have lots and lots of resources thrown at them.

Microsoft SharePoint search promises 'one size fits all', CMS included, and collaboration part of the deal. Do you believe this type of pitch and promise? What do you recommend to an organization that is unable to winnow the goose feathers from the giblets?

Microsoft SharePoint is a very good department-level/small enterprise solution. At present it does not scale in performance or price, but that does not mean to say it will not. The problem with SharePoint at present is that the search roadmap is not at all clear, with Search Express at one end, the core search application failing to be anywhere near best-of-breed (good news for third party vendors!) and FAST Search at the very top of the functionality pyramid in terms of cost and implementation complexity.

The police raided Microsoft Fast. Fast allegedly fudged the books because revenue did not support the numbers. What does this tell us about the viability of most search companies dependent on licensing fees and consulting services? Are other search companies in a similar pickle barrel?

Search requires a substantial R&D effort if the vendor is to stay ahead of the game. That means a lot of clever people and a lot of sunk cost. There are tough times ahead!

Search is now free. Download Lucene and you have search. What is the impact of open source software on commercial vendors? Does open source search technology have a place in an organization? How does an organization figure out whether to go open source or go for fee?

Compared to the CMS market (and indeed just about every software market) open source solutions are a very small element of the market. Lucene/Solr/Flax do things well enough to stop others challenging them but the R&D effort needed to develop technologies such as entity extraction and semantic search do not fit well with the open source model of distributed development.

There are more than 300 search, content processing, and text analytics vendors. How does an organization decide what to license?

Use a consultant who tracks the market to support the procurement process. Care needs to be taken to understand how companies such as Gartner and Forrester categorize the high-flyers in their reports. These analyses should just be one input. The CMS Watch reports are another source but it is also very important for organizations to network to find out the implementation experiences of others. The Intranet Benchmark Forum is a good example.

I don't think most organizations know what they don't know. Does your book provide guidelines for a company to identify what it doesn't know before it makes a decision or do you just tell companies what to do to procure a system?

I think that aspect is covered in the book but now that you have asked the question I think I’m going to go back and check!

What are the three must-do steps in the search procurement and management process? What is the risk of not doing one of these steps?

Find out what and how people search/invest in a search team ahead of investing in the technology/ make a huge effort to track changes in user requirements and expectations.

How can a reader get a copy of your new book?

Go to the Galatea web site. The book can be downloaded or purchased as a hard copy.

Can you give me a three sentence summary of what is in the book and why I should buy it?

I’ve spent over 30 years trying to understand search, I have a solid background in information science, I’ve helped companies select search applications and successfully implement them and I track what is going on in the industry. And I’ve made a lot of mistakes and learned from many wiser people than I. The book might not provide all the answers but I hope that it does raise all the questions that organizations should ask before, during and after implementing enterprise search.

    Navigation bar  
ArnoldIT Home Articles Speeches Services Features About ArnoldIT Articles Speeches Services Features About ArnoldIT ArnoldIT Home