Barney Pell's Weblog

Chris: First, Integrating product search throughout the network. Most people start
their experience at a search engine like Yahoo, Google. We work to integrate
that, like a shopping tab on yahoo search, and integrate relevant shopping
results directly within yahoo search results. If you're looking for a movie
within Yahoo movies, you can click directly over to the product purchase and
comparison page. Second, increasing the tools to make it easier for people
to shop. Save product results, share the list with other people. It helps
bring people back to the site. Community is mixed in with that, share lists
and reviews, and personalize. Will bring people back to shopping.

Michael: Search engine is as good as the results that it generates. If
someone comes along with a better search result and experience, then people
will flock to it. Page rank won't be the dominant ranking algoirthm. We came
up with our own algorithm, affinity indexed ranking (AIR) that delivers more
targeted results for shopping. And dynamic suggestion that shows you
different suggestions as you type in your query. Gets you to think about
things you may not have thought about. Continue to innovate and come up
with a differentiated feature.

o Gary: That's like expanding the query. Is there an equal challenge to
narrow the query, e.g. which kind of massage chair?

Michael: You want to go broad as well as focused. With vertical you can go
broader within one domain.
IBM used to do everything. Then Intel comes along with a processor, MSFT
comes with OS, and that completely shifts the balance of power. Something
similar may happen in the search area, where genearl search engine does
everything, from shopping, travel, etc all kinds of search. Vertical search
engines come out with better results, and people will migrate to them. The
total cumulative traffic on vertical sites may be bigger than on general
search engines.

Gary: Vertical also means a better interface to the same index?

Michael: Yes, like swiss army knife vs power drill.

o Gary: How things get into our index. Crawling, feeds, etc.

Graham: You can crawl as many sites as you want, but that's just half of
it. The second half is putting products into the relevant categories.
e.g. "Grill", different types, like bbq grill, car grill. For these types of
products, it's not easy to put them into the right bucket. It's easiest when
you have the manufacturer part # and product name. But reality is that
merchants don't cooperate. You need to go through them one by one and put
into the proper categories. From a merchant standpoint, if someone is
clicking through on a grill for cars and get to a bbq, that's not a
qualified click through and will hurt conversion rate.

Michael: We took a different approach. Launched our search engine with a
crawl search first. We crawled entire web and indexed just the shopping
related pages. Merchants don't give you all this information like product
reviews and buying guides in their merchant data feeds. We think there's a
neat for crawl search engine over unstructured data, and for a comparison
shopping engine based on structured data from merchants. There is room and
need for both for consumers looking for right info before they buy.

Gary: So you'd get a feed from merchant X and also crawl sites without
feeds. A merchant complained that all the feeds looked different. What's the
horizon for her?

Chris: Our strategy is to embrace both crawl and feeds. Feeds are useful for
high quality data. Crawl is useful for long tail query, obscure items. We
have algorithms that take crawled data and apply structure to that. It helps
to increase breadth of results we can return.

As for standardization in feed format, we agree and think we should work with the other shopping
sites for this. No conversations on this toipc yet.

q: Sam Perry, ConsumerSearch helps convert browsers into customers. Payment
for performance vs clickthrougs, do you see any shifts in the payscale for
people who are prepared to purchase, since they're more qualified?

Michael: Google has 46% coverage rate on adsense. That makes sense since
some queries have no sponsored ads. On become.com we're getting 95%
coverage, as people are looking for products to purchase. We heard Google
is getting about 5% click through rate on sponsored links, and we're getting
twice that. So that's a clear demonstration that it's much higher quality
traffic.

Gary: Part of the pitch to retailer is: "don't buy keywords on search
engines, buy them through us as they will click at a more qualified
rate". Do you have merchants reluctant to share?

Graham: People come to pricegrabber to buy. But on google, when they search
"grill" they may be looking for recipes. So people should be more qualified.

Gary: And if you sell only one type of grill, not another, Pricegrabber can
bid on the category word with more confidence.

q: How reliant are you on the big 3 for your advertising revenues?

Chris: Yahoo Network is of course a key part of our strategy. We don't
advertise off network, but look for more ways to integrate with the Yahoo
network over time. The other thing we do is keeping users on the site,
beomcing more engaged and qualified to buy.

Graham: we buy keywords. We also put resources to our cobrand teams. But the
biggest percent of results to our merchants come from people who come
directly to pricegrabber.

Gary: Ebay made a point they're only spending 15% of their buying budget on
keywords, not saying we're being overly dependent.

Michael: We're not dependent on any bought traffic right now. We aren't
doing search engine advertising. We'll do some search engine marketing when
we launch comaprison shopping. But it will be a small portion as we have our
own algorithmic search engine that delivers quality results.

q: For buying cycle, how what % are purchasing on your site, vs going
offline?

Chris: There is a large number of people that research online and buy at a
local store, that's the way most people buy. We have another vertical
search, local/mobile search, that helps address that. Users will print sheet
out for store and get a map. We also have a clipping service where you can
get product information via SMS.

gary: When people purchase offline, are they doing research on products,
retailer or both?

chris: You can get user reviews, consumer reports data on our site. The
second thing they want is a map.

gary: Separating out on and offline vs online-only merchants, to make the
interface fit the way people are using it?

Michael: When we launch comparison shopping, we intend to have local
inventory, tell consumer that this product is available online for this
price, but there's a store X miles away at a different price.
On the research site, we don't care if the user buys online or offline as we
just want people to come to our site for the research.

Graham: If we could get feeds from merchants (pricegrabber is more feed
driven), one of the biggest problem is merchants pay for traffic but don't
have the product in stock. If we could get inventory information that would
be great, but it's not that easy to get.
Second, there was a store that sells shoes in malls, and also have an online
property. There was a point when the two were listing identical dat feeds,
and the store that's only online they were getting great conversion
rate. But the offline store was getting significantly less conversion rate.

q: Using feeds, how do you categorizing the products?

Graham: We do have a team that categorizes, but we also have a taxonomy for
merchants to categorize the data themselves. Part of our company is going
through these products and putting them in the right places. Out of 1000
items, we categorize less than our competition. If someone categorizes our
product just as "shoes" we won't even list it. You can't buy a more
qualified shopper than on comaprison shopping.

Gary: you put some of the responsibility on the merchants...
Is the taxonomy pretty static, and if a merchant wants to make a change,
can they do that?

Graham: we don't require people to use our exact taxonomy. But if the
products are popular enough, it will get done fast. We both want to get a
click.

q: How do you handle relevance vs merchant's interest in getting user attention?

chris: Relevance is key for users, and we have algorithms to support
that. We also have speical programs to let merchants include their logo, fold in
their names, without compromising relevance.

q: Opportunity for someone to start a tagging operation?

graham: People can definitely do that now. We don't treat taxonomy or data
feed specs as company jewels. But as long as we can bring the most
compelling product to our users, then people can catch up. We have built
trusted relationships with the merchants. They have new comparison shopping
guys calling all the time, but hard to break through.

q: What's next, what about things like RSS, persistent searches, new ways people are
accessing information now and in the future?

Chris: Using standards to open up our data. E.g. RSS feeds for top
products. E.g. new computer releases have an rss feed.

Graham: We have 300+ cobrands. We're providing top products in RSS
formats. Most clicked on grill in pricegrabber in last 14 days. So our
programmers can put "top 5" links on each category.

Michael: We plan to support RSS feeds on our search engines so people can
keep tags on results. Also plan to launch become.com shopping news feed
service. Stay informed about product recalls, etc.