Web Support Blog

February 27, 2007

Writing good titles for searching and browsing your site is important

Filed under: Content Effectiveness, Search — Chris @ 11:21 pm

A few months back I did some work to assemble a training for our knowledge base writers on how to write good titles. Why? Well no matter how good your content may be, if your users cannot find the content, they wont read it. Take for example if you were reading a list of search results — you will determine the relevance of the content (whether you will click-through to read it) predominately by the title. The title is often given more weight to enterprise search engines as well.

Often the case is that you do not have a bad search engine, you just have bad titles. Elinor Mills of CNet wrote an article this month that provides some good examples of poor titles from newspapers who post their stories online.  Here’s a great example from the article:

Caught speedingNew York Post on Barry Bonds’ amphetamine use (2007)

I believe this illustrates the principle well. Very simply, if you were searching on Barry Bonds because folks in the office were talking about potential drug issues, it would be highly unlikely that this title would compel you to click through. Jakob Nielsen also wrote an article about this way back in 1998; not much has changed since then.

If you search Google with “writing good titles” you will find several articles, for which you could come up with a standard for your organization (Jakob’s 1998 article was the 3rd result). Take for example “Writing Better Page Titles” (2nd result); though it refers to web page titles, you can apply the concepts to knowledge content titles too.

Let me conclude with three tips that are important to me:

  1. Make sure the most descriptive and important words are within the first 3 to 5 words of your title. For example, if you had several titles on “How to do…”, it would be difficult for the user to find the specific how to they want. For this article, I have “writing”, “titles”, and “searching” within the first 5 words.
  2. Write your title last. It is okay to have a temporary title to use as your objective or guide while writing, but go back to it and refine once you have finished writing your content.
  3. Search your site with the concepts or phrases that you would expect users to use to find your content, before you write it. Perhaps you will find there is already similar information available, and you just need to improve on it. Or perhaps the information returned is very different, so you may need to rethink how you would write your title (and content) before you publish it. Finally after published, repeat your search to validate your content is returned as you would expect.

How are the titles on your web support site?

January 27, 2007

Outsource systems, so you can put more focus on innovation

Filed under: General, Search, Web Analytics — Chris @ 10:51 pm

In an article by Jon Brodkin of NetworkWorld.com, he discusses a recent comment from Google’s Dave Girouard on how “insane” it is that companies spend 75% to 80% of their budget maintaining, instead of innovating. If you consider a company like Google, or even Amazon.com and eBay, they used innovative technologies to drive their businesses. So why don’t others learn from that?

In my current organization, we have solved some of this issue through outsourcing, while other systems have issues because there are too many owners for different parts within our organization. First let me share two short examples of successful outsourcing – a web analytics tool and a customer survey tool.

We use Omniture’s SiteCatalyst as our tool of choice to perform web analytics. Omniture takes care of the data collection and management – an area that has been a problem in organizations that I have been a part of in the past. In my last organization, log files needed to be joined from the various proxy servers and made available to a WebTrends installation. Occasionally the IT organization would change something in the proxy servers, and we might lose a day or two of data (until the problem was found and resolved).

Of course we are not completely free of responsibility with SiteCatalyst. We are responsible to tag our web pages, which then communicate web activity back to the Omniture environment. If we should make a mistake in our tagging, or forget to add tags to a new web page, then we would miss that data. The other thing we do is send additional data, via a batch load, to Omniture that further classifies our data. For example, the page tag will send a product id, then through the batch load, we will send a product name, which makes it much easier when creating and viewing reports.

The other outsourced example, our survey tool, is similar to Omniture SiteCatalyst, as the vendor manages the data for us. In both cases, we free up resources that would otherwise have to manage the data and the servers that host it. Of course in both cases, there are times where we want to join that data with data from other systems. Depending on the complexity, we can either load data into the vendor system, or extract it from the vendor system to ours. We typically would only do that for ad-hoc data analysis, as routine reporting has the required data transfers automated.

So that’s the good news. But when it comes to search, it is quite the opposite. First, you could outsource your search engine, your entire web site, or perhaps just the knowledge management portion. For example, RightNow Technologies provides the knowledge base infrastructure for companies such as IBM, Nvidia, and RealNetworks.

In my organization, and in many others, search itself takes many different specialized skills to make it work.

  1. Managing the network and security access
  2. Maintaining the OS on the servers
  3. Maintaining the search application, including indexing, and its custom code
  4. Develop and maintaining the user interface, including passing the appropriate data to the engine
  5. Develop and manage the ontology and dictionary rules
  6. Content maintenance (e.g. have the appropriate meta tags)

If we were to outsource, we could have someone else worry about the first three (1 – 3), but the last three (4 – 6) will always require people from our organization (4 could be outsourced, if we outsourced the web site development and maintenance). The missing element on the list is someone to have oversight; someone to keep all the various parties in contact with each other. The same resource would also be required if you outsource, though again the first three would be replaced my managing a vendor relationship.

So whether we are talking about outsourcing security, as Mr. Brodkin mentions in his article, or applications within your environment, your company can free up resources for innovation, while paying experts to maintain a system for you.

November 9, 2006

When Is Search Not Search?

Filed under: Search — Chris @ 11:41 am

Usability guru Dr. Jakob Nielsen tells us that our users have a firm mental model on search. The mental model includes a search box, a search button, and a list of results. In testing, people want search on every site to work like search on their favorite search engine (Google, Yahoo, and MSN).

Users also expect to type in a keyword or two. “In [a] recent search study, the mean query length was 2.0 words.” Search success rates also drop off drastically as users refine, from 51% on their first attempt, to 32% and 18% on their second and third attempts, respectively.

In the knowledge management and enterprise search space, most vendors are pushing natural language searches – completely different than the mental model Dr. Nielsen describes. Nielsen recommends that if that is the case, then do not make your search look like search, “…the label ‘Search’ equals keyword searching, not other types of search.”

Dr. Nielsen does not have a hard, fast rule to solving this problem, but he does recommend not using a “Search” button. Try instead using the word “Find” or “Retrieve.”

So, “When Is Search Not Search?” When you do not support keyword search.

October 30, 2006

Components of Enterprise Search

Filed under: Search — Chris @ 4:03 pm

Do you ever wonder what the parts of a search engine are? Or, how would you evaluate search engines across enterprise search vendors? Well, if you visit the InQuira Resources (http://www.inquira.com/resources_reports.asp) you can order a whitepaper, “InQuira Self-Service and Support Search Solution – Advanced Linguistics, Dynamic Navigation and Classification” that can help you. The paper is an analysis done by the Patricia Seybold Group specifically about InQuira’s search solution, but it has information that can help you no matter who the vendor is.

The Patricia Seybold Group has developed a six category evaluation matrix that ranges from retrieval and results management to architecture to company viability. The whitepaper reviews each of these categories, many of which have up to five sub-categories.

In addition, the paper describes a search effectiveness ladder. The idea is that the higher the engine is on the ladder, the better it is. Keep in mind that the assumption is that natural language search is a good thing. Most of us are very familiar with Google, which is keyword search, and its algorithms are much different than an enterprise search tool. So assuming you buy into natural language search and you are willing to training your users on how to use it, here is the ladder:

  1. Keyword or text search: The more the word is mentioned in the content, the more relevant the content is scored.
  2. Natural language processing (NLP): Recognizes grammar, concepts, and relationships between words.
  3. Add synonyms to keyword and text searches (e.g smudge = smear)
  4. Advanced NLP: Introduces new relationships between words such as “contains,” “is part of,” and “occurs with.” Also identifies parts of speech and mines, classifies, and matches concepts.
  5. Intent: Classifies a question to make it actionable. The intent category is linked to language rules, ontology, and user experience categories.

So whether you are looking at a natural language search engine or a keyword search engine, I think this document is a good reference to use when analyzing all the components requiring consideration when purchasing a search solution. Of course if you already have a search solution, most of these categorize still work to make sure you have put the effort into each area.

October 15, 2006

When to Focus on Findability (Instead of Content)

Filed under: General, Search — Chris @ 12:06 pm

Recently I heard someone say that content is king – and I agree. But he said content is king, and that is where you should focus the majority of your attention on your site. To take that at face value would be a mistake. If you do not have any content on your site, of course you need to establish a process for creating and publishing content – a topic all on its own. The point I want to make today is that no matter how much content you have, if you cannot find your content, you might as well have not invested the resource to create it.

Take this a step further – you can focus your resources to develop high quality content, which can be expensive, and yet if your customers are not finding the content, you will have invested precious resources on developing content that no one is using. Consider your success rate for users ability to find your content – if it is 50% of the time, then one of every two content items will be found.

Here is how I look at this. Of all your content, how many have been viewed within the last month (or quarter). Do not confuse this with your success rate – if you had 100 content items, with 50% success, yet only 20 items were viewed, you potentially have an 80% findability (search or browse) failure. Of course since you only missed 50% of the time, your problem is no more than 50% — the rest is unnecessary content.

I will break this down further. Begin with this question: how many of the failed 50% could have been solved with the 80% that were not viewed? Assuming another 30 content items could have solved problems, then you have a 60% failure (and a 50% of unnecessary content). Therefore only 40% of your content is being found.

Findability1

Realistically you probably do not have all the answers for all of your customer needs. So assuming you could solve another 25% of your customer queries if your customers could find the content; and assume it took another 10 content items, then you would only have a 30% failure (10/30), with 70% of your content being waste.

Findability2

So we can see that putting the majority of our focus on content is not always the right approach. Instead, figure out how successful you site is doing with findability, and use it to drive your investment. With this new information, there are some strategies for dealing with this. Typically if you do a good job with findability, you can maintain that with minimal effort, so once you address these issues (search and browse), you can put most of your effort on content.

Perhaps you do not have the skills to address your findability issues, well… due to findability being more of a one-time effort (and then monitor for issues), you can conceivably outsource much of the work. This is especially true for search. There are many folks that can do great on site design and architecture, and you may already have one of them on staff, but search is a completely different issue.

In future posts, I will talk more specifically about search, and recommended approaches. For now if you look for outside help, here are some issues to explore: 1) keyword vs. natural language search (hint: Google uses keyword); 2) tagging vs. no tagging: balance resources for tagging vs. performance without tags.

September 22, 2006

Fingertip Knowledge: Learning-on-Demand

Filed under: Learning, Search, Web Analytics — Chris @ 5:48 am

I recently listened to a podcast from Elliot Masie, Fingertip Knowledge: Learning in a “Flatter” World. Mr. Masie introduces a concept called fingertip knowledge. He is using it in context of learning, but it is a concept I recognize — I have called it learning-on-demand. It is the idea that I wont try and learn all that I need to know, I will use Google to find what I need, when I need it.

This also translates to our support web sites. As I have indicated in my prior blogs, users will come to your site to learn a specific piece of information to solve a problem. Your users are employing the concept of fingertip knowledge more and more every day.

It is this idea as to why I suggest that the best way to determine success on your web support site is to look for the users who left your site immediately after reading a content (knowledge) item. Once you have fulfilled their learning-on-demand need, they will return to their work. Remember, as a provider of support information, customers come to your site because your product or service failed — once solved, they will go back to using your product to complete their real work.

Create a free website or blog at WordPress.com.