I have found on our site that approximately 105 of the visits to our site initiate a search as their primary vehicle for navigating around the site.

I guess that does not meet Filberts data, but I also believe that a users behavior is dictated by multiple circumstances.

- when they are searching
- the depth and complexity of the site
- their association with the site, is it first time or 20th time
- their understanding of the topic in questions
- the design of the site. Is the IA and nav similar to other popular site, or totally alien to the user.
- their previous experiences with search in general
- their web experience, I think experienced users convince themselves they can find anything without assistance.
- the urgency of finding information.

so when you take all these into account it kind of falls out that you simply need search, because you have no idea of the sate of mind the user is in when they get to your site and you really can't predict %ages of total traffic that will use search and that wont.

Posted by: JB on October 11, 2002 12:38 PM

---

 

105=10%

Posted by: JB on October 11, 2002 01:19 PM

---

 

Search Considered Harmful ?

When I bought the UIE Report last year, I was a bit miffed, as I had just spent three years building and maintaining a search and navigation system for our intranet (10,000+ users).

The system gives me real time stats and reporting, so I know that 68% of accessed links through the search, whilst the rest used the navigation categories. Last year I wrote to Jared to tell him that he was being harsh on search and we had a short discussion about it.

I then sulked for a while and decided to spend some time examining my data to see if it would help. I tried a few experiments, by making small changes to the system and watching the results.

I did not talk to users or get them to fill in any questionnaires, so these observations are only drawn from by hours of writing analysis programs and poring over a database that contained information about millions of searches and navigations.

I observed that:
Individuals used a mixture of search and navigation and there was only one group that seemed to rely on one more than another. These were new starters who spent a lot of time using the navigation. It seemed that they were browsing the navigation structure to get a feel for the organisation and the products it sold (financial products, mortgages etc). After a period of a few weeks they would then move onto the normal pattern of mostly searching with a bit of navigation.

Most of the searches (about 70%) were for items that the user have previously accessed through the search, and therfore knew that they existed. People would often use the same search dozens of times a day and always access the same result from the list. By looking at the searches performed it seemed that the users were lokking for the shortest word they could find to access their required result. Because the system has a real time usability feedback loop, it 'learns' from the users failed searches and updates its dictionary accordingly. This means that 'advanced' users learnt that the system seemed to 'know' their peculiar abbreviations and slang terms and would get them the correct result on demand.

The search and nav tool is implemented using the search pane of IE, so the users can performs searches without replacing the content in the browser window. This means that people could perform searches and look at the results before commiting to a navigation.

The most popular items in the system were the Daily News page and the Top Searches and Views page, which showed people what the most popular searches and navigation items were.

I think that people use the search most of the time, not because it is a traditional search, but because of the extra value they get from the 'slang' and 'abbreviation' short cuts, the ability to have a direct route into the 'improvement feedback loop' and just because it gives them the chance to be a voyeur.

I have spent the last six months trying to create navigation systems that don't rely on search but do their best to represent something about the strucuture of the domain that people wouldn't normally see. Hence webword stats and usabilityviews.

I am really starting to believe that 'plain search' is not a good thing, and misses out on an enormous amount of information.

Anyway, I could say more, but I had better leave it there for now.

PS. I really could do with some help on the graphic design for usabilityviews as my insane colour sense is beginning to take over, and I'm finding it hard to stop myself.

Posted by: Mac on October 11, 2002 01:23 PM

---