I came across these ads the other day and thought I would share them. These ran back in 1999/2000 when Snap.com was promoting itself heavily on TV. They are focusing on the fact that Snap.com used GlobalBrain technology to power their search.

Why is this relevant to site search? SLI Systems acquired the GlobalBrain technology back in 2001 and it is used in our Learning Search product.

We’re now powering the site search on Danny Sullivan’s Search Engine Land. Danny has requested a slightly different format to normal - with with the facets at the bottom. You can get to them by clicking the more options link beside the sorting options:

Most of our customers have the facets either at the top, or the left hand side. Those that want to conserve space will either use drop down for the facets or use java script to hide them (e.g. see Edwin Watts). Danny’s approach is different but not one I would recommend to most of our customers - almost no one uses the facets at the bottom of the search engine land page.

Search Engine Land is a relatively new site. Danny started it when he left Search Engine Watch late last year. Because it is new it doesn’t have that much content yet - but is getting bigger every day. It is interesting watching the top search terms. iCrossing has been popular - mainly because they were having a few issues with the company and Danny inadvertently gave people a forum to vent their frustrations in the Searcharazzi section. Disapointingly he shut it down - things were getting interesting.

Danny is the long time coordinator of the Search Engine Strategies show. Next week is the San Jose show. This is going to be the last one that Danny will be organizing. We will be exhibiting there and would love to see you if you’re at the show.

Google posted a video showing how Carolina Rustica (one of our customers) are using Google Adwords and other products - like Google Checkout. Richard Sexton said they shot over 8 hours of footage to make this 2 1/2 minute clip.

It’s a nice video - it puts a very human face onto ecommerce.

I spotted today that one of our customers, Ulta - has registered for an IPO. It’s great to see them doing so well and I would like to publicly offer them my congratulations.

I’ve been travelling for 4 out of the last 6 weeks and have neglected this blog for that time. I’m back now and should have a chance to do some posts soon. I’ve been to the ACCM show in Boston, the Internet Retailer Show in San Jose and I’ve been in holiday in Australia. The shows are a great opportunity to meet a lot of customers, prospects and partners in a short period of time. They are very stimulating and I’ll do separate posts on each of them.

I also enjoyed meeting some people that read this blog - including the CEO of one fo our competitors who said he finds it informative. More to come soon…

I would like to share with you my thoughts on how user reviews can improve site search on an ecommerce site. We are seeing an increasing number of our customers allowing their visitors to write reviews and rate their products. Many are using our partners, PowerReviews or BaazarVoice to do this. On a personal level, I can see why this trend is happening; I feel more comfortable buying a product if I can read reviews first. By offering reviews on an ecommerce site it means the prospective customer doesn’t have to go somewhere else to get these. Having reviews on a site should increase conversion rates.

The additional information collected from these reviews and ratings can be used to enhance site search in the following ways:

• The ratings summary (normally 1-5 stars) can be displayed in search. In particular this draws peoples attention to the products with good ratings.
  

• refine and sort by rating - this allows people to more easily find products with good ratings.
  

• Use ratings to weight results. The idea here is to give products that have a better rating more prominence in the search. You do need to be careful with this type of approach - just because a product has a good rating doesn’t mean it’s relevant for all keywords. But where two products have similar relevance it makes sense to show the one with the better rating first. Our initial experiments with this have been positive.
• Index additional text in reviews. The reviews add additional text to your product pages. This is useful for SEO purposes but it is also potentially useful for your site search. If the text in the reviews is indexed by your site search then your visitors are going to be able to find results using a broader range of phrases.

User reviews empower people by allowing them to express themselves and influence the choices of others. This fits nicely with SLI’s Learning search which uses peoples selection from search results to improve the relevance for others. Together user reviews and site search are a nice complement.

Last month I had a chance to talk to Aaron D’Souza from Google at the Search Engine Room conference. Aaron is a Software Engineer in the Search Quality group run by Matt Cutts. I asked Aaron about the new Google webmaster guideline:

• Use robots.txt to prevent crawling of search results pages or other auto-generated pages that don’t add much value for users coming from search engines.

In particular I asked how Google determines whether pages “add much value for users”. Aaron’s response was that if the pages are ranking in Google then they are deemed to add value for users. I was pleased to hear this.

I described how we were concerned about a customer of ours that had over a hundred thousand search and navigation pages indexed by google even though they had less than 5000 products. We were concerned that Google may penalise them. Aaron reassured me that this was fine and suggested we use the sitemaps feature to indicate to Google which of those pages we considered the most important.

Martin Kelly just sent me the photo which prompted me to write this. He also sent this one of me giving a presentation.

I saw Jacob Nielsen discussing the importance of breadcrumb trails and how they should be implemented.

We use the breadcrumb trail in our search and navigation solutions. It reflects the search that was done and any restrictions or facets that the user had selected. In the example below I searched for My Chemical Romance on Hot Topic, then selected Apparel then Band T-Shirts. You can see the bread crumb trail at the top reflects those selections.

We follow most of the guidelines that Jacob recommended. Because the facets don’t have any natural hierarchy we show them in the order in which they are selected. On the few times we have tried showing them in any other order we have always had complaints.

We put together some interesting stats from across a handful of our clients on misspellings on internal search. We used a broad definition of a misspelling. If a search phrase contained a word which is not on the site then we count it as a misspelling. Obviously this will also include words that are spelled (or spelt?) correctly but just aren’t on the site. For example if you search for “razor” on Tupperware.com it’s not on the site anywhere so we would count that as a misspelling even though razor is spelled correctly. So this yields an upper limit.

The average was 10% of queries contained a word that wasn’t in the index. This was higher than I was expecting and emphasises the importance of having a spell checker on your site search and of paying attention to your search terms that have poor results (which is a larger set than the terms which contain words that aren’t in the index). Interestingly the client with the lowest rate of misspellings is one that pays a lot of attention to our report that shows search terms with poor results and regularly updates their synonyms to try and improve these.

Last week our sister company, Eurekster announced that they had received $5.5M of Venture Capital. SLI was one of the founders of Eurekster. We helped develop some of their initial technology and shared offices with them. They have been running without any serious input from us for a couple of years now and it’s great to see. They have had over 50,000 swickis created.

With the capital on board they’ve now got a chance of being very successful. I wish them the very best.

Note: I referred to them as our sister company. Brother company would probably be more accurate since my brother, Grant is the largest shareholder and Chief Scientist at Eurekster.