Lijit Alpha? – Google Custom Search Engine Is Not Currently Suitable For Site Search

Lijit Search

Lijit

There has been some talk today about Lijit, a service intended to be a better way to search for information. I was pointed in the direction of it by a post by Matthew Chen of Megite, which in turn led me to another blog I subscribe to but hadn’t looked at yet, A VC, who discussed a Lijit search for Venture Capital blogs that is now appearing on the VC Feedburner Network coordinated by Brad Feld who also discusses it.
This is all based upon a partnership between Lijit and Feedburner. I hadn’t seen anything on the Feedburner blog about this partnership but as I am interested in Google CSE and Feedburner, I thought I would take a look.

Setting Up Lijit

It is quite easy to setup an account with Lijit, so I am not going to concentrate too much on it. The most important thing I need to mention is you should create a new Google Account, never use your main Google Account especialy if you use any sensitive services… hmm every Google service that uses IDs is sensitive – dont use your normal account.

They couldn’t find my Flickr account, which does exist but doesn’t have any photos yet
They needed my delicious account details – I gave in on that one, no financial info there

Glaring holes

I couldn’t use MyBlogLog data (though maybe an API is needed, and I know they are working on one)
Adding OPML wasn’t obvious when I was performing my tests, thus these tests are based on without opml

My Lijit Network

Firstly it should be noted I don’t use bookmarking services extensively, but I have digged a few SEO posts in the past on various domains such as seobook, searchengineland, seoblackhat etc

From my own site I have also written a lot about SEO, and in particular linking of various types, so I thought that would be a good search term to try.

I have a few SEOs in my LinkedIn, but I have no idea how they access who I know on Linkedin

Lijit Results

If the intention is to provide some kind of comprehensive site search of your own site, plus all the sites you have some kind of association with, I am not impressed.

Search Term: Linking

My Content:-

Lijit My Content

Looks like they are missing out on a lot of content, and don’t know what my site is as there are results from other domains as well. only 2 of the 5 results were on my domain. I will discuss the limitations of Google CSE later.

Search Term: Linking

My Network:-

Lijit My Network

Five results again – actually exactly the same ones – there seems to be a major bug and the search only returns “My Network” which appears to be minuscule.

It appears that the only “network” results are specific pages that contain the term, rather than searching the domains you have digged, bookmarked, stumbled in the past. Quite frankly – useless to the average user.

It also unfortunately suffers from the limitations of Google CSE – it doesn’t pick up all pages – in fact very few. I am fairly sure that Google CSE currently only shows pages with a lot of authority, and thus needs a large dataset to return good results.

Search Term: site:andybeard.eu Linking

SEO Search by Alister Cameron

SEO Search linking

Again not many results, but at least they are my tag pages, that will lead to relevant content /tag/linking and /tag/linking_strategy

Someone with high pagerank and lots of inbound links is going to be easily impressed with Lijit, because a search will show most of their content. The ideal candidates to demo this with? A lot of VCs with authoritative blogs. Their pages might show up in the results, so it will look cool. A great target would be VCs that use a lot of social bookmarks.
For any normal user, with a domain pagerank of 5 or less, it is useless for finding content.

SEO Search does bring up quality results for SEO and SEM from 350 sites, but only the very best content. Maybe still a little harsh, and maybe Google are not updating the data as fast as the main index.
You can also get Toolbar Buttons for it (want one for VC Search?)

Search Term: site:andybeard.eu Linking

Google Search

I thought it was important to demonstrate how many pages on my site talk about linking.

Google Search Linking

That is a bit more realistic, 526 pages – there might only be 20 or 30 pages of unique content that have that word, but that is the power of tagging in creating lots of variations of pages ;) Technorati do the same.

How to Fix Lijit

It is quite simple to fix Lijit so it gives some relevant results

  • Forget about the money – Lijit should use Google Free web search or do a deal to use Google Adsense for Search – they will provide much better results from a small data set
  • Add Whole Domains – if you have bookmarked a site once, include all the data in the search from the domain, not just a single document
  • Fix bugs – I understand it is beta, but I couldn’t see a huge amount of evidence that the additional network sites I added were even included in the results I saw – maybe it should be called “alpha” – even the OPML file I later added after my tests was not used
  • Income sharing – Google CSE can be used for large networks, but revenue sharing would make it a lot more attractive

At the end of the day, no one is going to use a search that doesn’t work, thus it is best to use something that works, and worry about the monetization version later, when it can give results.
In the games industry, a game had to work, and be 90% complete to be called a beta. I would even expect more usability from an alpha.
Google need to make CSE a little more forgiving than it currently is. Google CSE is a beta and just needs some tweaking for certain applications.
I don’t think Lijit is quite ready for implementing on production sites other than those of high ranking VCs. Once it really reaches beta, it might be promising.

Liked this post? Follow this blog to get more. Follow

Comments

  1. says

    Andy – thanks for the direct feedback and for putting Lijit through it’s paces. It’s definitely early in the dev cycle at this point and the search dynamics are improving daily as the guys at Lijit tweak and tune it. I’ve forwarded this on to Stan James (Lijit CTO) and I’m sure he’ll take a good look at the “How to Fix Lijit” suggestions.

  2. says

    Andy, thanks for looking at Lijit. We are working day and night to make results more relevant in all situations. We agree its way to hard to add OPML files at this point (although it can be done). Look for changes in the next week. We are committed to working with Google as a partner and we have had good discussions with them over the last week concerning some of the issues you mention.