Live search comparison

Tonight I was thinking about the importance of the number of indexed weblogs when choosing a search service. PubSub claims to track 6.6 million sources, Technorati claims to track 4.7 million weblogs, and Feedster claims to track 1.1 million feeds. What can the target audience of these services do with this information? I decided to do some comparative research from the point of view of a marketing department tracking the buzz around their new advertising campaign.

Kevin Kringle is a digital word-of-mouth marketing campaign created by Best Buy and SMG Reverb. The marketing campaign is officially under a month old so it should be a good proving ground for live searches. The search phrase is “Kevin Kringle” in all cases.

22,500 results
Last mention 5 days ago
10,300 results
21 results
Last mention 3 days ago
8 results
Last mention 5 days ago
2 results
Last mention 11 days ago

Still waiting on results from PubSub. Feedster’s results page shows 100 results even though there are only 21. Feedster phrase search did not display correctly on my results pages.

The big search engines (Google, Yahoo!) performed better than I thought they would although you cannot yet subscribe to a search from a big search engine. Despite errors presenting the data, Feedster appeared to have the most results and most recently updated listings. Technorati and Feedster could both do a better job with on-page promotions of persistent search capabilities.

  • Posted
  • Updated at
  • Comments [6]


Commentary on "Live search comparison":

  1. alex on wrote:

    I got 11 on jyte, although 2 of the headlines point to the same webpage.

  2. David Sifry on wrote:


    Interesting comparison! We currently don’t show links or posts that no longer on the front page of a site, so that inherently puts Technorati at something of a disadvantage. As such, Technorati is natively biased towards conversations occurring in the past 2 weeks or so. Not that we throw the data away, we don’t – but perhaps we should allow you to see older, archived posts as well in the search results. Good food for thought.

    Also, you’re missing all of the people who link to the site:

    which currenntly shows 12 links from 12 sources on Technorati at the time of this comment.

    Shows 15 posts matching at this time as well.

    BTW, ever try to see all of the “22,500 results” that Google claims it has? Good luck!

    Anyway, thanks for the great test – clearly we still have some work to do to get even more complete coverage…


  3. Niall Kennedy on wrote:

    I wanted to track the penetration of a marketing idea which may or may not include a link to the Web site. Kryptonite comes to mind as a meme a marketer would like to track but few links to the Kryptonite Web site.

    As a new user the natural inclination would be to search by phrase, the same behavior as most people are used to with Google. Most users do not know about “link:” or “linkdomain:” to track their Yahoo! inbound links. When the Procter and Gamble employee comes to Technorati’s site my guess is they will search for Downy instead of

  4. Bob Wyman on wrote:

    Comparing PubSub to search services can be a bit of a nuisance since you are often comparing “apples to oranges”. They are similar in that they are both fruits, but otherwise very dissimilar…

    PubSub only does prospective search, not retrospective search. PubSub results accumulate over time rather than appearing the moment you key in your query. What I often do when making these comparisions is start by creating a PubSub subscription, let it run for a few days, and then run retrospective searches on Google, Feedster, etc. to see if they found the same, more, or fewer items then PubSub did. The alternative, is to start all the tests at the same moment, but I’ve found that doing that doesn’t really improve the accuracy of the tests — as long as the test period is less than a week or so.

    NOTE: It is very important when constructing your PubSub test query to make sure you aren’t unintentionally filtering out some results. If you just use the default search box on our home page, you’ll be getting contents from more than weblogs and you’ll also find that the weblog results are limited to those in the more frequently linked to blogs (based on LinkRanks calculations.) To ensure that you are searching across all of the weblogs we monitor — and only the weblogs, you should create a “focused subscription” (see right side of home page) and ensure that you check the box for “Include High-Volume Weblogs and Personal Journals” and DO NOT check the box for “Limit results to the top n% of sources”.

    Also, note that while we we monitor 6,664,709 feeds, we consider many of them to be inactive. We only claim to monitor 3,628,145 active feeds at this time. We keep watching the inactive feeds since occaisionally one of them will “wake up,” however, the vast majority of them seem to stay inactive once they get in that state.

    bob wyman

  5. Niall Kennedy on wrote:


    I created a PubSub subscription for “Kevin Kringle” when I authored the post but this posting is currently the only result.

    I compare total sources indexed because Technorati does not offer an active comparison, and if they did it would most likely not be consistent with the PubSub definition of an active source.

    Eventually you are looking to make a sale, or capture an extra user to generate ad revenue. When I see past results it is a much more compelling sell than waiting for what could happen in the future.

  6. Bob Wyman on wrote:

    I just created test entries containing the phrase “Kevin Kringle” on my test blogs at LiveJournal and AOL. Both were found via PubSub almost immediately — neither has appeared at either Feedster or Technorati after over an hour. Feedster’s most recent hit on “Kevin Kringle” was your post which it seems to have found five copies of 1 day and 19 hours ago. Technorati’s most recent hit was also your post and it found it 10 hours and 52 minutes ago — much later then Feedster did, but thankfully Technorati only found one copy. BlogLines only has two “Kevin Kringle” hits now – the most recent was found on Nov 12. BlogPulse’s most recent hit was two days ago… Google doesn’t do much better. Only PubSub shows my two posts at this point. It will be interesting to see how long it takes before my posts show up on the other systems. But, right now, I’m tired of waiting. I’ll check again in the morning…

    I think this demonstrates what I would have expected: i.e. PubSub is a superior system if “prospective search” is what you are looking for. If monitoring the broad and dynamic blogosphere is what you need, then use PubSub. If researching the past is your goal, then Technorati, Feedster and Google are all good candidates to use for your “retrospective search”. They do something completely different from what PubSub does. I use all three in combination with PubSub. The key is to know which to use and when.

    You wrote: “past results [are] much more compelling sell than waiting for what could happen in the future” Yes, for many users that is the case. However, there is a distinct class of users who require what we provide: Real-time, personal matching across the largest number of blogs. We realize that this is only a subset of the overall market, however, it is a difficult challenge to satisfy these needs since there has been so little academic and commercial development in Internet Scale matching of the type we do. On the other hand, the technology needed for retrospective search is well understood and has been for quite some time. That’s why you’ve got so many people doing it and why it is so hard to build a distinguished service. We’ll get around to doing retrospective search eventually — either by building it ourselves (I’ve done it a number of times in the past) or by partnering with someone else. (Anyone want to provide our retrospective search backend? Send me mail!) In the meantime, we’re doing our best to build the first, and the best, Internet Scale prospective search system… I hope you find it useful.

    bob wyman