Thanks again for a great product. We recommend it to anyone who have a business website.

Michael

Michael

I would like to know how do we implement internal search within website.

Thanks and Regards

You need to know your site’s internal search data. What’s internal search data? This data is a gold mine. If you know what folks are searching for, you can tailor your site to their need. I’ve seen conversion rates jump by as much as 20% when tailor your site to internal search phrases.

Google’s new Custom Search Business Edition is a great search engine for small to mid-size sites. But it doesn’t provide much in the way of useful analytics - a little graph of search volume, and a short list of heavily searched phrases, if you’re lucky.

Enter Google Analytics. You can use Google Analytics filters to create detailed, easy-to-read onsite search rep….

Coming from the user experience and information retrieval worlds, I haven’t had many opportunities to interact with people from the analytics community, especially over this topic. For the most part, the advice here is in accord with what we’d see from folks with my kind of background, although many of the comments here suggest that the end goal is a transaction (which isn’t surprising for analytics people). In sites that are more focused on content and research, there generally aren’t quantitative measures to go on, so search analytics more typically supports qualitative evaluations.

One issue that I don’t remember seeing here: using search analytics as a diagnostic tool to improve such areas as:
* Interface design (both of the search interface and how results are presented)
* Content tuning (identify content gaps, improve existing content to ensure that it’s high on the results list)
* Organizational behavior (search analytics become a driver for getting disparate parties involved in a web site to the same table, and can even have an impact on organizational strategy)

So much to cover; wish there was more time (and room) here… In any case, thanks again; I hope this discussion will continue!

The second half of #3 is both interestng and unnecessary. :-)
Why completely spider your own content? You know what it’s in there, it’s your own database!

Why not build a list of changed/removed pages or similar, only spider those pages and do a delta add to your SE.

Much faster, more efficient and can be done often.

We use this method with verity and hence, our internal search engine is at most 30 minutes out of step with the site content.

We could be quicker, but 30 minutes seems a reasonable balance all round for our needs.

Steve

1. What is the average number of searches per visit? Search solution provider Monosoft states that an average of 2.4 searches occur per visit. A ratio above this benchmark could be a sign that your content is too heavy or your navigation structure needs work.

2. What is the percentage of visitors going directly to search from their initial entry page? Depending on the kind of site, a high number might be good or might be bad. For exameple, on a retail site a high percentage of new visitors going directly to search is probably not optimal. For a “library” or research site, a high number might be desirable among repeat visitors.

3. How many searches resulted in “0 matches” as a percent of search attempts? You want this number to be as low as possible. Misspellings, the use of phrases associated with a competitor, or even lack of your search solution crawling all your content can be the culprit. You would be surprised how often, for various technical reasons, the entire site is not crawled.

4. What percent of visitors gave up on your search and abandoned your site from the search page, and what is the average number of searches this takes? For added information, correlate this with the average number of search results returned for each search. These metrics address such issues as these: Is your search facility returning too few results? Too many? Are they “scan-able” and easily understood?

These 4 should be enough to get most people started on optimizing on-site search…

Allow me to butt in with a couple of related usability considerations:

If you use Click Density / Site Overlay you already know that it is difficult to find a feature (or even the dreaded “path”) that is used by 10% of your site traffic. Yet internal search is a feature that is being used by 10% or more of your site traffic.

That’s the reason why I usually recommend adding the top 10 search key phrases in a content box adjacent to the Search box itself, making it more visible.

I see this as a way to circumvent/minimize the obvious usability issue: your visitors do not want to spend time looking for the Search box, or as you suggest, they fail miserably while looking for it :-(

Also, keep in mind that tools such as Click Density and GA Overlay will not work correctly with content viewed on PDAs, phones, WebTV.

I wish I could have replied to your comments in a timely fashion today, here is a catch all set of comments (right around midnight in the US : ):

Ann : Your point is well taken. But take your example to the extreme and imagine all these support websites running around with rather sub optimal internal site search engines and analysis. There you are not only costing annoyance but real tangible $$$ taken away from the bottom line. Sad.

Gradiva : Thanks for putting the pointer to Louis’s upcoming book. After I finished the post last night I did drop a note to Louis requesting him to check out the blog post and share his feedback with all of us. I hope he will.

(For those of you interested in SEO I want to put in a plug for Gradiva & Jennifer’s book: Search Engine Optimization : An Hour a Day, four full stars on amazon! : ))

Jen : Both synonyms and best bets are great features to leverage from your search tool but then throw in Click Density / Site Overlay on top of those results to validate that people are clicking on your synonyms and best bet suggestions. If not rinse and repeat. : )

Justin : The reports you suggest are great, you have already made more progress than the rest of us.

I am so glad that you are striving to add to rest of your site data. That can be super powerful:

• Internal keywords by site referring urls or major sources of traffic etc.
• Tying to Conversion Rates (as Steve above also suggested).
• Funnel analysis for those who do internal searches vs those who don’t (specially powerful on support sites)

These are just some of the benefits. Would love to see the demo, am sending you a email.

Benry : On the link you provided to your blog the most delightful thing was the emphasis of looking for “Failed Searches”. From my humble experience as well I’ll stand completely behind that, it is really amazing how much insight we can get for stuff our visitors can’t find.

In one famous example one of the top five keywords consistently was “register”, when the company made it explicit in the box and during software install and in other ways that the product can’t be registered via the web. The customers still came looking and there was no content on the company site about it.

Jeff : I am somewhat disappointed that you are only “somewhat” of a fan. Just kidding! : ) Thanks for your input, it is always welcome.

I’m not here to plug, but I really like Omniture’s ability to break down user migration patterns based on values associated with their custom variables and the ability to correlate each with one another - as well as applying segmentation through Discover, which is something my company has found very useful the last 6 months.

My only gripe is the cap of unique values given for each variable, which I discovered recently was applied just last year (November) which has put slight wrench in our solution. Nonetheless, I find the tool great for the needs of our end-users.

Anyhow, great content… I look forward to reading / learning more.

Jeff

I’ve rigged up some reports in HBX that deliver all of this to me daily using Report Builder and talk about it here if anyone is interested.

http://www.benry.net/blog/2006/04/07/scent-from-internal-search-terms/

[Source: Occam’s Razor by Avinash Kaushik] quoted: Life Lesson: You can put the greatest tool on earth, Google, on your website but if your websites are sub-optimal (url structures, content key words, missing meta tags and best bets) the…

In the Urchin case are you able to integrate GSA data with the core clickstream data of the site?

To answer your question, no. At least not yet ;)

The module is meant to process the logs from the device in a standalone fashion. I do not tie the reports back to other log sources because there is no clean way to piece the user session together between the log files.

With that said, I do create a number of useful reports including:

Traffic Reports:
—————-
Traffic Summary
Sessions Graph
Searches Graph
IP Addresses
Depth of Sessions
Length of Sessions

Search Reports:
—————
Search Terms
Keyword Progressions
Keyword To and From
Exit Keywords
Search Term Number of Results
Zero Result Searches

If you want to see a demo drop me an email and I’ll send you a link.

Justin

We don’t just focus on the top 25, but actually look thru the entire list. Adding common misspellings to the correction pool.
We also check things like searches that return null results, and pick the more common searches and random others to see just what they do return.

We do occaisionally see results that are odd or unexpected and hence indicate need for further tuning.

It is certainly not a static set and forget part of a website, but a highly critical dynamic that needs just as much effort as other parts of the site.

We do find that it does tie nicely into our tracking from external SE’s. Mainly Google. This gives us additional clues as to what visitors are looking for at a broad level for the more generic terms. Google’s Trends is an absolute goldmine for this.

Fascinating stuff!

Steve

Thanks,
Marshall

http://www.webmetricsguru.com/2006/06/_site_search_analysis_you_shou.html

The second is what Manoj mentioned: measuring search terms that get no results is absolutely key! (Another great one is search terms for which no links were clicked.) A client I work with discovered that a top term was a mispelling of calendar. Clearly they were looking for the calendar, so they added a synonym. That may be a bad example since some search engines know how to handle common mispellings, but this client uses their no results terms very heavily. They know that users certainly expect relevant results from site search.

Gradiva)