The "Official" Verdict on Powerset (or what would I do if i was the CEO).

Powerset has received a ton of attention–some say hype–about their “Google-killing” intelligence based search. Google-killing is in quotes there for a reason people: no one is killing Google any time soon, and in fact the best search and navigation companies–i believe–will learn to play in Google’s pond. (note the “official” in the title is also a joke, but I’m sure some folks will miss it–the same folks who will miss me explicitly saying it’s a joke here–thus is blogging: fire, ready, AIM!).

Anyway, this week Powerset finally, after close to two years of work, released their first product (which I really a demo truth be told): a site-based search engine for Wikipedia. Last year at the TechCrunch40, in September, they showed an earlier version of the product.

How does Powerset stack up?

  • When evaluated on the basis of a site-based search engine for Wikipedia it’s fantastic because Wikipedia’s native search sucks (like really sucks). Powerset is amazing for *site* search.
  • When evaluated as a search engine to compete with Google or Yahoo, however, Powerset’s current offering is a failure. Indexing a couple of million pages of Wikipedia really doesn’t do much for end users, and when you compare queries on Powerset head to head with Google or Yahoo it’s embarrassing.

Of course, the point of this demo is *NOT* to compare PFW (“Powerset for Wikipedia”) to Google or Yahoo today, but rather compare it to two things:

  1. the current search engine at Wikipedia (which is, again, a disaster)
  2. a site specific search on Google or Yahoo (by adding the operator site:wikipedia.org to any search you do on Powerset).
  3. a normal Google or Yahoo search (without the site:wikipedia.org operator).

Let’s take a simple question, “who created the ipod,” and run it against these three shall we? Now, there is a simple answer to this question, “Apple,” which no one would have to search for. Let’s assume this question means we want the name of the person, or people, who were responsible for making it.

In the Powerset search we get EPIC FAIL: The first result is for software made to work with the iPod, and the results below it are not helpful.

In the second scenario when we search Google with the site:wikipedia.org operator we get FAIL! Obviously looking for the question in a system filled with statements is not going work.

In example three below we see that a straight up Google result indexes another search engine–the one known as Yahoo Answers–which has the right answer mixed up inside their results. NOT FAIL! Of course, is this victory Google’s or Yahoo Answers’? Most would agree it’s Yahoo’s because they have questions in their database, not keywords.

The most interesting point, for me however, is not Powerset’s search feature, which has much to work out, but rather their navigation tool. IT NAILED IT! EPIC SUCCESS!!!

Clearly Powerset has some value, however it is not going to be as a front end for user search. Users do not want something this complex for 90% of their searching. For 90% of their searching they want something very, very simple like Google, Yahoo, Ask, etc.

Folks do NOT want to type a sentence into a box, rather they want to type one or two words into a box and get a huge payoff.

The problem, of course, is that in our business you need to not only make a neat piece of technology but you also have to build a brand and match, or exceed, users expectations. In it’s current form Powerset is not going to do that.

If I was the CEO of Powerset I would do one of two things:

  1. Partner up with an existing search company and provide a secondary search option. For example, if you do a search and you come up with a result from Wikipedia you would give an extended abstract (say 2-3x the normal length). So, imagine if you searched on Yahoo or Google for the quesiton above and in the results you had the sentence:
    • As ordered by CEO Steve Jobs, Apple’s hardware engineering chief Jon Rubinstein assembled a team of engineers to design the iPod, including hardware engineers Tony Fadell and Michael Dhuey, and design engineer Jonathan Ive.
  2. Become a business to business provider of search solutions. If Wikipedia won’t pay for this kind of search, I’m sure the CIA, FBI, IBM, educational institutions, financial shops, etc. will.

Anyway, congratulations to the PowerSet team on burning the midnight oil for so long trying to make search better. As a company we’re struggling with similar issues (i.e. the partial data problem: we only have 45,000 guide pages built), and the fact is that search is a huge problem that is not going to be solved by one company. Search is a problem that will take a couple of dozen PowerSet, Hakia, and Mahalo type companies.

I wish PowerSet a lot of luck and can’t wait to see what happens when they index ten sites!

Challenge for the JasonNation: Pick your own question to ask PowerSet, Google, and Yahoo, then put your results in a similar fashion (i.e. with the screen shots) on your blog. JasonNation members who experiment with the results and do a post will get one of the sexy new Mahalo mugs (version 2.0).