The SBS Interview: Brian Gil, Yahoo Local

In the days after the SMX West conference last month, I recapped the “Local Search & Blended Results” session here on SBS, pointing out a comment from the Yahoo! speaker about the non-impact of reviews and ratings on the three local business listings that show up in Yahoo!’s Local Shortcut.

Problem is, I got the comment wrong. 🙂

That Yahoo! speaker — Brian Gil, Director of Product Management for Yahoo! Local — then clarified that the ratings are a factor, but not the review text. And in the process of getting all that clarified, I was invited to speak at greater length with Brian about Yahoo! Local, local search tactics, reviews and ratings, and much more. We spoke on the phone Tuesday, and this is the full text of our conversation.

MATT: This started after the SMX West presentation when a bunch of us who write about local search jumped all over your comments about reviews and ratings, and that being a tactic for small businesses trying to get found in local search. I know that you’re not going to reveal the whole “secret sauce,” but I was hoping we could talk more about that and some of the other things that go into the local search mix.

BRIAN: I know there was a little bit of back and forth about that on the blogs. But the core thing is to ensure that the base data that we have in the business record are accurate; things like the title of the business and its category.

But one thing I was also trying to stress in that presentation at SMX was about going after the long tail of queries. I think that’s an area where small merchants, even more than chains, can stand out in local search and really build their business. And that’s by providing the depth of content that’s available to them in our self-service listing program. They can really lay out all the products and services and brands they carry, write a long description — all that content does get indexed and does factor in to what we determine to show.

Ratings and reviews is certainly a contributor, but it’s not the end-all, be-all. And I was trying to clarify that rating and review content from a customer isn’t enough, on its own, to let the business qualify to show up for a certain query in the first place. It’s more about, once the listing qualifies to show up for a query, what we call “recall” — the listing is recalled by the search engine — the rating is then taken into account when we determine the sort order. But it’s more critical to make sure that you are part of that initial recall so that you have an opportunity show up in the 1, 2, 3 spots on search, or the top 10 spots that show up on the first page within Local.

I always encourage business owners to just make sure that their business records are as robust as possible and to clearly lay out what differentiating services they provide. Because we are seeing — as consumers are getting more savvy in how they look for local merchants online — we’re seeing more complex queries, certainly a shift away from the traditional heading and business name-type lookup queries that you see in IYP (Internet yellow pages) sites. More people are using Web search-style queries. Maybe it’s the size of the query box, I’m not sure yet, but people tend to be more descriptive in what they’re looking for. It’s a huge opportunity for merchants to rank very well without trying to compete in the old-school way of, ‘What’s my category? It’s plumbing, and therefore I have to rank for “plumbing”.’ It’s not just about that anymore. It’s much more diverse than that.

You mentioned the idea of qualifying for the initial recall. A business can have a listing in Yahoo! Local with a lot of data, but then there’s also the business Web site. Does the Web site also play into the initial recall, or is it just the Yahoo! Local business listing?

That’s one where, depending on the context in which we’re using the content, we might choose to use it differently. So, if we’re in a context which requires higher precision, and therefore could opt to have less recall, we wouldn’t include the Web text that we glean from their official Web site or any other Web pages that come in. So I would say, in general, no, that is not enough for a listing to show up in the first place.

What I mean is, if a user is on the Local property, and we’re not coming up with extremely precise results to fill a query, we can include information like that to make sure we come up with an answer. But the confidence in that answer on our end would be lower, and we’re careful about when we opt to recall those types of listings.

The context dependency is key, because in properties such as Web search, the user is looking for an answer to a question and there are many different routes they could go. They could go the direct listing route that Yahoo! Local would feed into there, or they could browse through the algorithmic results from other Web sites. And we’re very careful about what we choose to put on the top of that Web search results page. It’s all about precision, and making sure the user is directed to the most appropriate answer at any given time. But if the user is self-selected into a vertical already, they’ve kind of already been very specific about their intent, and we can be a little more liberal with the recall.

I want to get more into business data in a moment, but let me ask a couple more things about reviews. Is there such a thing as “power reviewer”? In other words, is a review from someone with a history of writing reviews more authoritative than someone who just left his first review?

No, not from an organic perspective, but that’s where the social and community elements come into play. The users who do opt to identify themselves with a name, they upload a photo, and they do a repeat number of transactions — that user is going to have more credibility among the community. So with the more qualitative analyses that we’ve done, where we bring people into the lab and we show them businesses — some of those businesses are unreviewed, some of them have reviews, but from anonymous users, and then others are reviews with pictures, where you get more of a sense of who the reviewer is. Definitely the latter is the most influential, so it is important to have that human connection.

That’s why, last August, we made some product changes to our review process to really promote quality and the connection between users. We eliminated the ability for users to leave a rating and not submit a review. We also eliminated the option to post reviews anonymously. And we added a minimum character count to the reviews, as well.

Those things were in response to what users were telling is in the lab, and what merchants were telling us from their perspective on fairness with these types of services. They know their reputation is on the line, and we heard from various merchants … ‘Hey, someone being able to go to the site and with one click give me one star and then provide no context is almost unfair.’ They feel like they’re not empowered to understand the rating, let alone do something about it. So on that request, we added in the comments feature, where business owners can comment back on reviews, and either agree with or disagree with the review. Often we’ve seen business owners go in and say, ‘We’re sorry you had a bad experience. Come in again and identify yourself and we’ll make it right.’ And that’s exactly the kind of dialog that we want to encourage.

One last question on reviews: People like myself and others who write about local search on the blogs — do we make too big a deal out of reviews and their impact on the local SERPs?

At this point, I think so. Yes. I would focus on that long tail and matching, and not worry so much about reviews. I know it’s almost starting to play on people’s psyche, because they say, ‘Oh, no, my competitor has four stars and I only have three!’ And while it does ultimately matter, I wouldn’t say, from a ranking perspective, that it’s where the majority of effort should be focused.

Let’s get back to the idea of business data. There are all sorts of sources of data about a business. Can you speak to the importance of that business data being consistent? For example, if they change phone numbers, and two data sources have the wrong number still, but a third source has the correct number — how much does that mess things up for the business owner?

It certainly can. A lot of of our efforts are around understanding the quality of the sources and the freshness of the data that comes from the various providers that we obtain data from, and we blend all that together. It’s certainly a challenge. We have 17 million records in the database and they are in flux continuously. We’ve done a very complex bit of logic for who we trust, literally at the field level — for each field of content. We do look at changes over time, and who says what, and how many concurring sources we have. It is a challenge.

But the good news is, if the business owner claims their listing, they don’t have to worry about those other sources being outdated. If they take ownership, they can control it. But I can understand the complexity, because there is no one single clearinghouse that a business owner could go to and trust that the whole Web will be updated. But if they take the time and — at least for us, I can say definitively — if they take the time to actively claim it, they won’t have to worry about it going forward.

If I use Yahoo.com and search for [financial planner in Portland], I’ll get a Local Shortcut. But if I search for [financial planner Portland], I won’t get the Shortcut. What are some of the factors that determine when a Local Shortcut appears out on the main search engine?

I can’t run through them all off the top of my head, but there are certain trigger terms — like “in” — that help clarify intent. There’s also the prominence and uniqueness of the location that’s inserted, so that we can clearly understand the location as a modifier to the query. Someone typing [portland trail blazers] does not mean they want to trail blaze in Portland, Oregon. And, of course, if they include the state code, such as [financial planner portland or], that would’ve worked.

It’s all about a confidence, about knowing that we have the user’s intent accurate and can show local results at the expense of showing algorithmic Web results. If we don’t reach that confidence threshold, we tend to be conservative and will opt to not [show the Local Shortcut]. Local results can show up in the algorithmic results, so the user can get their answer that way, too.

It’s all about ensuring that we put that best possible answer at the top of the page. There are some complexities there, but the more specific they are with the location — a zip code would trigger it, a state code would trigger it. If the city was more unique, too — maybe the issue was Portland, Oregon, versus Portland, Maine — I’m not sure. But it does vary quite a bit, and there’s not a hand-picked list or anything. It’s all rule-based.

I think it was sometime last year [actually, it was late 2006] that you started allowing users to modify and add data to business listings. How has that worked out so far? Have there been problems? Has it been well accepted? And then part two of that question is, did you see the article yesterday on Search Engine Land about the hotel affiliates taking over the URLs for the actual hotels?

On the first point, we’ve been very pleased with our Consumer Submit program. I’m most happy with two things: First, the addition of new businesses in the community, that the community hears about first, way before it goes into the local phone book. Consumers are coming in and adding [listings] on their own accord, which is great. The other thing is that businesses close, and we like to get information about that from consumers, too.

They can also make edits, and with edits and closures, what we do with it once it goes through our moderation is to treat those as comments on the business. We don’t actually overwrite the listings that we have, but the listing would have a notation saying, there’s some changes suggested by the community, or it could say, this business may be closed. The intent is that the user can call ahead to check it out. When business listings are added to by consumers, we’ll put a note on the page so that, again, the user may want to call ahead before they jump in the car.

So, yeah, I’ve been very, very pleased by that and we haven’t seen what I would categorize as significant abuse issues. I’m not going to speak specifically to the hotel thing. That one is a unique case. We have been looking into it. I don’t want to say that it’s necessarily as nefarious as it was written up. In fact, a lot of businesses like to use URLs that are redirects for tracking purposes. Whether or not that click-through is for an affiliate tracker to use to get paid is a bit that’s in question, I would say. But if people feed content to us, we can give them the option to display a URL for tracking it separately. That’s often the purpose behind that, in addition to the reports that we give our partners on usage.

So, we’re checking in to it. We’ll take the appropriate action, but my gut is telling me that it’s not nearly as suspect as what was written up.

There is a danger, though, when you open up the system to allowing modifications by users. Certainly you guys were aware when you did this that the risk was there.

Yes. Whenever we roll out a service like that, we always couple it with increased fraud prevention measures. We don’t just throw more people at it, because oftentimes a person couldn’t even detect a pattern of spam. So, we always complement our human moderation team with automated controls, some of which will automatically reject submissions if it determines that something’s gotten out of hand. Or, it will just raise [an issue] as needing deeper investigation by human beings. All of that is in play, and we were very cognizant of the risk.

Couple last questions for you, Brian. First, what’s the future of Yahoo! Yellow Pages? It seems to me quite redundant to Yahoo! Local. Is that going to change at all, or will it continue as it is?

For the foreseeable future, I would say Yahoo! Yellow Pages will continue to operate as is. It is a product which has a strong, core base of users. It’s a largely at-work usage, with tons of administrative assistant-types using it for lookups and efficiency. Its precision is dialed up extremely high, so if you want just a phone number, it’s a very efficient tool for that. It also has a very strong brand. We feel that the products complement each other, with Local largely being a search-based product, and YP largely being a browse-based and lookup product. It’s also highly valued by our advertising partners. So, as I said, I foresee it staying up and serving a different audience than Local.

Okay, last question. Can you give a sneak preview of what’s to come in 2008 for Yahoo! Local?

Sure, the Yahoo! theme of “open” which was touched on at the SMX West conference will be extended to Local so that we can give a broader perspective on the Web’s local businesses. I would also say that the localization of the Yahoo! network will be a big theme, as well.

We know that Yahoo! Local as a property serves itself very well. It also serves the Web search user base very well. But there are other areas in Yahoo!, other strong starting points which would benefit from providing more of a localized view with more local content to their user base. We, as a team, are excited about that opportunity, as well. So, opening up and distributing out to the network is the future of where Yahoo! Local is going.

There’s quite a strong local element, in my mind, at least, to something like Flickr, and obviously Upcoming.org is all about local events. Can that be integrated more in the future?

Absolutely. Upcoming.org actually is in my group, as well, so we see the events experience as a natural extension to local businesses, and they actually complement each other very well. When you go to an event, you often have other local business needs around that event, and it’s also a natural browse-and-discovery-based experience. People don’t necessarily know exactly what they want to do, but they know it’s Friday, and they have to figure something out.

So, I definitely see Upcoming and the whole events space as interesting going forward. And then when you talk about browse-and-discovery, something that motivates people is images. And having such a great asset as Flickr is certainly something that we look to leverage more going forward. If you’re thinking about going to an event, I’m sure the Flickr stream of photos from past occurrences would be very compelling.

— end —

Great thanks to Brian Gil for the conversation, and to Yahoo! for helping arrange it.

Previous SBS Interviews:
The SBS Interview: Andrew Goodman
The SBS Interview: Lee Odden
The SBS Interview: Jon Glick