Loku CEO: Leveraging Data to Make Sense of Hyperlocal Search

Share this:

The idea of “big data”  is hot these days — particularly in regard to local. Harnessing the huge amount of information being collected about location and cross-referencing it against individual people’s behaviors is no small endeavor, but it holds the promise of creating really dynamic products that can deliver the right ad to the right person at the right time.

In 2010, private equity manager-turned-entrepreneur Dan Street launched a two-year quest to create a search engine capable of synthesizing and analyzing the vast landscape of hyperlocal information. Working in conjunction with researchers at the University of Texas, Street’s team built Loku, a hyperlocal information hub that draws on big data tools to make sense of local search through dynamic analysis and a clean presentation of hyperlocal content. The Austin-based company launched a beta version of its service last month.

Street Fight caught up with Street to discuss the problems with local search today and how distributive computing can open doors for hyperlocal entrepreneurs.

Can you give us some background on why “big data” is so hot right now, and how it applies to search?
When Google created its search engine, they were trying to find the single-best web page as a result of a user typing in a search term. As a result of having to index the Internet, the team came across some technical challenges – mainly, the problem of processing a gigantic amount of data in a relatively short amount of time.

So, over the years, Yahoo developed competing technology and spun that tech out to be open sourced. And as that happened, a lot of other technology spun out to support it. In effect, the technology that Google built is now available to anyone who is smart enough to know how to use it. “Big data” is a term that applies to all of those tools put together.

What problem can big data solve in hyperlocal?
Local is a pretty tough beast. Let’s say you were looking to find something out about the honey badger. Search works great because there are a couple really good sources on the Internet that solve that problem. The problem with local is that, say, you’re trying to find a really good restaurant, the content sources are so distributed — you have a bunch of people talking about that restaurant in a bunch of different places — that you have to pull it all together and try to make meaning out of it.

When you have the ability to put together a huge amount of data and analyze that data on the cloud, you can do some really cool stuff. One of the things you can do is really understand language patterns, and really begin to understand what language means.

Why does local need a different framework for discovery?
Local is not a search problem; it is less about trying to find something, and more about synthesizing and understanding information. Since that is the constraint of the problem, you really have no other option but to try to create a solution that does analysis and synthesis rather than one that trying to use standard search to find the best result.

Local is not a search problem; it is less about trying to find something, and more about synthesizing and understanding information.

Before, when you couldn’t do distributive computing, you could have put together a bunch of servers to get it done — but it would have cost a bunch of money. Using big data allows you to go through and create meaning by synthesizing and analyzing all this information.  That’s the kind of thing that search can’t solve but big data can.

At the Street Fight Summit, Fwix CEO Darian Shirazi argued that LSO, or location search optimization, would be the next step in an ongoing process of web optimization. How much does geotagging content play in optimizing local search?
[Geotagging] is a very big part of it — maybe, 30% of the problem. When you have such a large amount of information, it’s really tough to figure out where this content is being applied or where it matters. And yes, this is a significant part of the problem, but it is not the majority of the problem. Even if you find out where that piece of content applies, you still have to figure out what that piece of content means.

So walk us through how Loku works in the back-end.
Let’s say that we find a bunch of information about a bar in Brooklyn.  The first thing we do is search for terms that might apply to that restaurant or bar. Then we try to figure out which of those pieces of content — social media, check-ins, blogs, newspapers etc. — apply to that place.  Once we bring together all of that information and understanding, we look for words, or patterns of words, that might indicate what this place is all about.

One of things we’ve found is that if you look at grammar patterns or word composition, it tells you a lot about people who like to go to a place. For example, the word “rad” is often used to describe places that are artistic or cultural.

But frankly, at this point these are rough-cut algorithms. Most of what we are doing today is manually priming this data. We take all of this information and tell it what that place is like in human terms, and then the machines take over and try to figure out what other places are like, using those human terms. We’ll take a place in Brooklyn, for example, and rate it ourselves, and we say “this place is hipster.” Than we go to San Francisco, and finds a similar place with similar language patterns and our technology says, “this place is hipster.”

Is an advertising-centered revenue model an effective way of monetizing a product like Loku?
I tried advertising for a little while, but the problem is that with advertising you’re never going to cover the sales cost.  Local businesses are willing to pay 100 to 150 bucks to pay for an ad.  But just the cost of getting those sales efforts off the ground and controlling for the amount of area you need to cover — you need like 10 people per city to establish really good coverage — will make you bankrupt. So our strategy is to partner with people in hyperlocal areas, who already have those sales forces. As an example, if a local entrepreneur wants to create a local sales force in Brooklyn he can sell ads, he can sell deals; we can just help syndicate those and take a marketing cut on the top. We’re trying to sit on top of the other hyperlocal media, not compete with them.

Where do you see the hyperlocal space heading over the next two to three years?
I’m excited about the opportunity for local entrepreneurship and I think we’ll see a lot of very innovative ideas emerging from the space.  But I think there needs to be a substantive mind-shift from the traditional model in which, we’re looking for highly scalable, gigantic businesses, that go after huge opportunities.  The vast majority of hyperlocal will center on smaller, more entrepreneurial, lifestyle businesses. I think its going to be an ecosystem with a lot of startup-like business popping up in places like Brooklyn and Boulder and a couple of bigger companies, which are pulling together the various innovations and trying to make sense of it all.

Steven Jacobs is an associate editor at Street Fight. This interview has been edited for length and clarity.