That's pretty interesting

Thanks

Wow, what a few months. A wedding, a trip to Africa, and a tumultuous return to work that included a last-minute invitation to sit on a panel at the semantic web conference.Unrelated to all of this I’……

Many recs initiatives fail because they either (1) try to build relationships between users instead of static objects or (2) try to build relationships between static objects that don't show strong enough intent. Amazon's recs work because it's algorithm ties products to other products based on purchases, a very strong signal of intent. The problem with (1) is that cold start takes too long – you force users to answer dozens of questions before they get any results, a less than ideal user experience. The problem with (2) is that using data derived from views, clicks, ratings, etc is very noisy, which causes you to have low quality recommendations. Many of the times Amazon has been chastised for poor recommendations have been a result of “Customers who viewed this also viewed”, which are only shown as a last resort.

Geo companies who require manual check-in have great intent — someone took the time to check in. So as long as your dataset is sufficiently dense, you can start to create “customers who check-in here also checked-in to” relationships, and from there building personalized recommendations: “Because you checked-in here, foursquare recommends: “. Since your relationships are between check-ins and not users, as soon as a new user checks in any where, you can immediately recommend them similar check-in locations, avoiding forcing them to answer dozens of questions and train a system.

All to say, I think geo is the first platform since e-commerce that can actually do recommendations really well. Exciting!

I also note that “SimpleGeo and Mixer Labs are pursuing what sounds like the Holy Grail — providing a one-stop shop that makes it much easier for developers to access and integrate a wide range of location building blocks into their mobile apps.”

Did I get the 2nd point right?

Phil Hendrix, immr

Here are a few observations that I think relate to this “stack”:
(1) Mobile phones are morphing into devices that connect users to places, other individuals and location-specific content in new and compelling ways…enabling a whole new class of location-centered services, many of which rely on visual capabilities.
(2) The ability to scan bar codes – both 1D and 2d versions, such as QR codes – with camera phones will revolutionize the way in which individuals obtain information, shop and more generally experience places.
(3) Geotagging is becoming increasingly automated, increasing the amount and accessibility of posts, images and other user-generated content. Automatic geotagging will (i) rapidly increase the frequency with which users add “location” to their social data; (ii) dramatically expand the volume of location-specific information produced; and (iii) intensify the need and create significant new opportunities for solutions that help individuals filter, find, access and leverage timely, location-specific content.
(4) User-generated content, particularly when geotagged and searchable by location, represents an enormous, largely untapped asset, which is why Twitter, Facebook and other social media companies are investing heavily to make UGC accessible. As geotagging of content becomes easier and more common, more “locally relevant” content will be available. As more content is geotagged more precisely, the accuracy of “nearby results” is also likely to improve, further reinforcing and boosting usage. Proximity searching and filtering results by location will prove indispensable.

Curious how these technologies/interfaces fit into your notion of stacks?

Phil Hendrix, http://www.immr.org

Personally I think we are moving from vertical to horizontal for two reasons: (1) vertical geo offers with all the layers Chris describes above are actually technically difficult to do (more on this another time); and (2) the existing geo business models don't yet work unless you're at very large scale.

Add to this the dynamics that Google, the largest scale player, purposefully runs their geo business at a loss and is actively destroying other geo revenue models (e.g. Citysearch & Garmin).

It takes a lot of firepower to build something non-trivial when revenues really don't happen until you are huge.

The missing piece is an organized market – one of the greatest opportunities in local monetization.

For your comment regarding expensive call center overhead to connect with SMBs, the only way to build long term, trusted relationships with local business owners is a local approach, not a remote boiler room.

We have started to break down the barriers for local, small businesses in the third party gift card distribution, empowering them to sell their gift cards and discount cards in national retailers in those retail stores nearby their business. We could not have done it without a local presence.

Our next frontier is to enable “flow” in the online distribution for SMBs and we utilize our local presence in each market to gather the content and freely have it distributed by developers, publishers and LBS through our “Deal API”.

Furthermore, there are diminishing returns to this. Being the first plumber
in your neighborhood with a strong online presence will almost certainly
give your business a big boost. Being the 12th — not so much.

Re: will creepy guys/stalkers see this, that's an excellent point. My wife
(who uses Foursquare even more than I do) thinks this is a problem for
Foursquare adoption vis-à-vis women — women are understandably much more
wary of broadcasting their location than men. I pester her to write a post
about this on my blog.

As far as Facebook is concerned, maybe I'm wrong, but I think the only
people who've cared about their privacy settings changes are us techies. At
least, it doesn't seem to have dented their growth/engagement significantly.

In any case, you're absolutely right that we have to look at the cost as
well as the benefit side of the equation. But I was responding to Chris's
contention that gaming elements and such probably won't appeal to Normals,
which I disagree with.

Partially I'm just excited for vast improvements to the “local web” everywhere. I'm amazed how few local businesses (plumbers, barbers, non-restaurants, etc) have much of a web presence that allows them to be found via Google and have basic, current info about services. I think there is a business opportunity there, but not sure how do to the sales to reach local, small businesses who haven't historically seen the value of the web.

However, unlike Techies, Normals really care about the *costs* of the cost-benefit equation:

- “Am I going to look weird or uncool doing this?”
- “Will my information be safe? Can creepy guys/stalkers see my info somehow? Will they be like FB and start making my stuff public in a sneaky way?”
- “What if they say it's free and try to trick me later and charge for it?”
- “What if some of my friends who are annoying start showing up where I am? I might feel bad defriending them.”

These are just a few of the questions that many people will care about. I am not saying that checking in won't have enough momentum and value overall but anytime you are changing social behaviors it's by no means a given.

(BTW, you clearly get these issues given your great Tumblr vs. Posterous analysis

To me, the problem in crowd sourcing of data that links to a specific business or place is that totally unstructured data is very difficult to aggregate, use in analytics or apply inference against.

Applications that sit on top of utility sets can drive MUCH more value than “starting from scratch”. As to how valuable the venue layer will be in the long term? I'm skeptical, if Google opens up its Place Pages into it's API, it flattens the upside for venue database providers. Will it be a key enabler, definitely, will it carry a lot of revenue value, I'm not so convinced.

Also, this is what they said about Facebook and Twitter (and even Google once) and they're all trending older, less tech-savvy, etc.

It's one thing to not being able to cross over from the Techies to the Normals (aka the FriendFeed Syndrome), but I think crossing over from young/urban/educated/blah to the broader population is a lot more trivial once you get to a certain size. Young people talk to their parents and to their bosses. They get the New York Times hot and bothered about their latest toy. Etc.

As you once said yourself, trends that happen in Japan/Korea happen in the West after a while, and people there have been able to pay for stuff at the cash register with their phones for a while now. The iPhone has Bluetooth; you don't even have to go through iTunes or AT&T, you can imagine an app that directly links your credit cards or bank accounts.

I think the “serendipitous meeting with friends” thing is absolutely what will make location based socnets appeal to Normals. When Foursquare first came to Paris, I had been clamoring for it on Twitter, nobody else had heard of it. During the first week, the Paris leaderboard was just my friends. And after my *very first check-in* I realized a friend of mine was in the same bar as me, on a different floor. That's powerful stuff. And I don't see how it doesn't apply to Normals.

I also think game mechanics could be huge for spurring broader adoption. It's a bit striking to me that you don't think this can be appealing to the general population, as this is something that Hunch does a lot. Also games are, I think, the biggest category in the app store. The normals love gaming, especially on mobile, and especially with friends. Foursquare is right at the sweet spot of this.

Also, tourism does seem like a very good early app for geo apps.

I think another layer of the geo/local stack is Real-time Activity. What is going on there (“formal” Events or “informal” happenings/vibe). This is an area we've done work in at Going where the main point of our mobile app is not just checking in but contributing a real-time rating, photo and comment on what is going on at the place. Buzzd and SocialGreat are doing work in this area as well (and I think Hashceratops is a great idea although as a former NLP'er the namespace issues are very challenging).

http://yellowarrow.net/v3/

I think Milo can play a crucial role not only in tracking conversion from online discovery to offline purchase, but also in the top value-added layer of the stack. What I mean is that Milo has enough real-time inventory data mapped to certain brick and mortar stores that could serve as a great utility. For example, I may have a to-do list of errands I need to run when I'm out. Whenever I'm closest to one of the stores carrying an item on the list, I can receive a ping that says “the toaster you needed to pick up is available 3 blocks away at retailer X.” I like the perspective of “geotility”– the idea that I place a higher value on information, goods, and services when they're useful to me in the right place at the right time.

I'm not sure if this is the right way to go about it, BUT there has to be some way to collectively organize everyone's hidden local tips/spots for a city.

With local services, part of transitioning to the “Normals” is also making something useful for those who don't live in NYC/SF. Much of the kind of serendipity that foursquare promotes likely has much less value outside cities.

With local services, part of transitioning to the “Normals” is also making something useful for those who don't live in NYC/SF. Much of the kind of serendipity that foursquare promotes likely has much less value outside cities.

The SP angle is to have APIs which give app developers access to location
and end-user preferences, etc from the network perspective. Perhaps another
layer of data to augment 4sq data.

claim is that it will allow developers to write differentiated apps which
work across SP networks in conjunction w/ APIs at the app level. Of course
SP pipes are getting commoditized quickly so this is their only answer to
traffic/revenue decoupling.

The SP angle is to have APIs which give app developers access to location
and end-user preferences, etc from the network perspective. Perhaps another
layer of data to augment 4sq data.

claim is that it will allow developers to write differentiated apps which
work across SP networks in conjunction w/ APIs at the app level. Of course
SP pipes are getting commoditized quickly so this is their only answer to
traffic/revenue decoupling.