Update – March 26, 2006: Dulance Dead?
–

From my phone interview with Sergei Burkov, Fouder and CEO of Dulance…

What is Dulance? How are you different than the other comparison shopping engines?
Dulance is a 2nd generation comparison shopping engine. The difference between us and 1st generation comparison engines is our breadth of coverage. 1st generation engines cover 5,000 – 50,000 merchants, and only those merchants that have established relationships through sending xml feeds. There’s a long tail which they are missing. There are many rare products which are not covered. So we went after the long tail, the rare products. Each rare product itself is negligible, but combined, the number of rare products is more than the number of popular products like digital cameras or electronics. Our goal is to cover everything and anything online.

We don’t rely on XML feeds, we spider the websites and extract the information using a form of AI. Essentially, it’s search that covers all online merchants in US and Canada. There are three parts to our technology:

1. Determine who the merchants are. The technology analyzes the content of the page and determines if the page is from an online store as opposed to a blog, review, etc. A shopping cart is one indicator, but we look for a lot of indicators and we have proper weights for different attributes in order to properly identify the online stores.

2. Price extraction. We know who sells a product online, but need to know how much it sells for. With a lot of scraping technology, you have to write a different program for each store. We developed a universal screen scraper that’s intelligent enough to find prices on all sites. We cover over 1,000,000 sites.

3. Rankings. When you search for something popular, there is going to be more than one store which offers the product. [The question is] which one to display first. Google’s PageRank is good for information, but not for shopping. Suppose you’re searching for a specific type of downhill ski. There might be only 20 shops selling that model. Page rank values the pages by how many other sites are linking to that page, but that doesn’t apply to the ski shop because that site doesn’t have many incoming links. However, the site is still very relevant. Different relevancies or rankings are required.

What about categorization? It seems to be a problem with crawler based engines…
It would be very nice to have everything categorized and [sites like] Shopping.com have done a good job categorizing products. At some point, we will add similar functionality for the best sellers. Unfortunately, though, this approach doesn’t scale. All the categorization has been done manually. When you [categorize] digital cameras, you need people who understand all the different attributes. You need experts in the product to figure out what attributes are important. If you can do it, you should do it.

The problem is that there are 100,000+ different product categories. How would you do the same normalization for [the less popular products] like gyroscopes, downhill skis, kites for kite surfing, etc. For kite surfing, you have to consider things like aspect rations. Kites have a unique set of attributes and to do a high level of classification, you need to be a kiter, it cannot be done algorithmically. For all 100,000 categories and 100s of millions of products, you can’t do the same thing Shopping.com does for digital cameras.

Nextag, Shopzilla, and others are helping bring the offline experience online.
Check Shopzilla. When products are normalized, you can search by tons of attributes. If you search for camping pants, though, it’s not normalized.

You still need something form the long tail. On Dulance, you type something in and you’ll get a slightly random result, but you go to a store and the store is very helpful in helping you make the purchase. If you go to a good kite store, [the store] will tell you the difference between 4 and 5 line kites. Before purchasing, you go back to Dulance. It’s not a linear process. Go to Dulance. Go read about the products. Find the model. Then go back to Dulance and type in the exact model you want.

How do you make money?
Through sponsored listings and affiliate links. We have a mix of CPA and CPC results.

Have you received financing?
We’d prefer not to discuss.

How do you expect to gain traffic?
1. Word of mouth. People will tell friends. Through guerilla marketing.
2. Syndication, partnerships with web properties that have traffic. We have signed up a number of partners. We can integrate our results into informational websites. Suppose you have product reviews…a site like Powder Magazine publishes a gear guide with ratings and descriptions every August or September. Today Power Magazine would just list the manufacturer’s suggested price. What Dulance can provide is a link which offers today’s prices. Readers click on the link and get the results showing who sells the product today and at what price. That’s a very valuable service for the readers. Early indications are that the click-through rate (CTR) is 30-40% compared to single digits for Google Adwords and other search advertising. Our results are super targeted. It’s great for the merchants as the user is much more qualified. Also, we are just talking about a text link which reads ‘Check Prices’ so it takes up less real estate on the page. The user is less annoyed.

How do you compete with all the other contextual offerings?
The difference is in targeting. AdSense decides algorithmically what to display. For Dulance, the webmaster has to highlight the product name and send it to us. This is why it’s so targeted. There are clear advantages to our product. In a gear guide you have a table listing 5 products or 5 different models. The webmaster could put up 5 ‘Check Price’ links and each one will give you the prices for each of the products. With AdSense, it’s impossible. Adsense will try to produce something that’s relevant, but by lumping all the products together, you’ll never have a good price comparison. Google AdSense and other services are good for blogs or magazines with vague content. But if you have a gear guide that reviews specific products such as a Linksys 4-Port Router, you should use Dulance. We’re best for product review sites.

The 2nd difference is in the way you display the results. There might be 150 stores selling a particular model. With AdSense, Google displays the ads for 8-10 of the stores and that takes a lot of space on the page. With Dulance, one click will give you those 150 results. It’s a more efficient use of advertising space and less spammy. When a user clicks, he gets what he asks for.

Other services…
We have a service called “Search the Future.” It’s comparison shopping through RSS. Essentially, it allows you to place limit orders on products. It’s useful for users who want catch sales/promotions at end of the season. For instance, sunglasses might be discounted, but then will be sold out in a week. What you can do is go to Dulance, type in the product, the price range, and subscribe to the RSS feed. We repeat the query on your behalf everyday and you get a notification.

It’s an interesting new approach because it automates comparison shopping and allows merchants to actually push their promotions or clearance sales. The RSS feed is a new channel of communication between sellers and buyers.

Is there room for 1st generation search?
There is room for 1st generation search and Dulance might accept feeds in the future. There are 10-20k merchants that provide feeds, but 200,000 that buy keywords, and 1m stores that are online. It will be a paradise when all stores have everything in a structured XML format. We’re not against the feed, and we’ll embrace it, but we can’t wait 10yrs for small ski shops in Vermont to give us XML feeds.

–
Related Posts:
Dulance Dead? – March 26, 2006