A number of writers in the mainstream press seem to think that Snap combines “organic” results with “paid” results into one result set due to some nefarious scheme to confuse searchers. Our supposed motivation is to obfuscate that a paid-for ad isn’t a paid-for ad, hence being able to get more advertisers at the expense of the trust of our users. This seems like a nonsensical approach if you want to be in business for long.

So, I’d like to take a few minutes to discuss why we combine the paid and unpaid results into one column. Here goes:

1. At Snap, we believe that every search has only one set of relevant answers that will – depending on the search – contain some mix of commercial and non-commercial results. For example, a search for an alternator for a 1997 Toyota Camry should contain almost all commercial results since the intent of this search is likely a purchase of an alternator. Conversely, if you want to research birth defects, our result set probably won’t contain many commercial results.

2. Next, let’s answer the question of how we combine organic and paid results together. No, we didn’t combine paid and unpaid in some random manner. Rather, we look at user actions to decide relevance. Because our pricing model is Pay-per-Action, we know how well our searchers “rate” a paid listing because our advertisers must report back to us when an action occurs (think sale, download, qualified lead, subscription, etc.) Hence, the more actions, the more effective a site must be in meeting users’ needs. For commercial searches, better satisfying user needs gets the advertiser farther up the search results.

We take a similar approach with non-paid results, but with a different set of data. We use the post click data stream from approximately 2 million searchers to determine which sites perform well on past searches for the same term. If a site consistently gets more user interaction, we assume it’s a more relevant site. We believe this is much more effective in delivering relevance than analyzing link strength. Because we’re using post-click behavior for both paid and unpaid, our algorithms can normalize for optimum relevance.

3. We clearly mark the difference between a sponsored result and a non-sponsored result in our listings set. This is what a sponsored result looks like in our index:

It’s true that we recently made the Sponsored Result label more noticeable, but that was because we didn’t believe it was that important to users in the first place. Why? Well the research we’ve seen (PEW Internet and American Life Project – Deborah Fallows, Jan 2005) indicates that the whole issue of paid/unpaid isn’t that big of a deal. We’ve now realized its importance, at least to the media, and have strengthened its visibility.

4. There are other benefits to combining paid and unpaid results. The primary one is that it allows us to provide users with a totally different interface – the visual display of results – that we think will deliver a better search experience. We’ve already written lots about this. But have you noticed how efficient this design is? We don’t have duplicate organic/paid results, we have at most seven paid results versus the 10 plus of traditional search engines, etc.

There’s an old saying along the lines of, “no good deed goes unpunished,” and I think we’re suffering a little bit of this effect as the sole reason we combined paid and unpaid was to provide better relevance. In our attempt to push the envelope, we apparently pushed the boundaries of what’s accepted in the world of search. Good.