Every day, 5 people lock themselves out in London

(and tweet about it)

Sylvain Carle // CTO & Co-Founder // @froginthevalleyIgnite Where 2.0 - April 19th 2011

Good evening. I am super happy to be back at Ignite Where.

Last year I talked about how people sucked at naming places.

This year I am going to talk about how stupid we are. And the fact that we brag about it with geo metadata to boot.

All my examples are from Needium, our platfrom that matches needs expressed to a location and to businesses that can answer them.

Image credits on each slide. Most are Creative Commons Share Alike from Flickr.com

10 geosocial patterns and

illustrations of the human nature

When I saw the call for speakers for Ignite Where, I wondered how I could talk about what we are doing with Needium without doing a blatant pitch or rant about how great we are.I hate it when people do that.

And while we donʼt quite have the clout of Google to declare the Zeitgeist I knew of a few interesting anecdotes about what people declare, and where.

So hereʼs 10 things we have observed about the patterns of needs expressed.But before that, I little background.

Lorem ipsum dolor adiscumFinally, someone who cares that you are

tweeting what you had for lunch! No really.

(and where, too)

So there is this Twitter stereotype, you know, the obvious reaction of a newbie when you try to explain to them what the frak is Twitter.

Of course, there IS a sliver of truth to every stereotype.

As a data geek, I was sure there was some value in there.

I always find that when there is a strong backlash to something new, itʼs a good thing, a sure sign.

http://www.google.com/search?q=who+cares+what+you+had+for+lunch

It all started when @sebprovencher told me he saw Twitter as this realtime stream of needs.Could we parse it?

And I work with this guy, Seb, heʼs just brilliant. I basically just try to keep up with all the ideas he has.Heʼs been thinking about local search for 10+ years.

He had this instinct about structuring local conversations, 4 years ago.We created Praized API to explore that. We kind of nailed important things, but failed badly at reselling our technology.Maybe we were too early. Or just really bad at sales.

http://twitter.com/sebprovencher

Anyhow, we did this site for asking a local question to your social network, based on this algorithms augmenting humans idea.

The idea was good, but we failed (again) to grasp one key issue. Why would someone go to a specific site to ask their question?

Couldnʼt they just ask it on Twitter and Facebook? In fact, we discovered they were doing that in droves.

But I mean, massively. We built a prototype in 2009 to asses how big this trend was. We were blown away.

Looking at Twitter as this firehose of needs is a completely different experience than looking thru the keyhole of who you follow.

I had a few doubts(to say the least)

Could we really @reply people “asking the universe” and make money while doing it?

So I had a few doubts. Could we really “just” @reply people asking questions. Could we scale this? Would they hate it?

We looked at the pattern around us and saw lots of status updates asking questions and recommendations.

Word of mouth ruled even before the internet was born. But now we had this incredible word-of-mouth machine.

Hey, maybe there was even a business to be built on this.

1. Build Cool Twitter Harvesters. 2. Do Something. 3. Profits!

Too easy to be true... (South Park reference)

Things not to mess up(important)

•It’s SOCIAL Media, no bots•Be nice, no HARD sell. Help.

After running the service in Montréal and Boston for a few months, we had a few key insights.

Itʼs social media. I am a bit scared, even, by social media experts. I hate spam. No auto-DM. Never.

It should be the incarnation of the Cluetrain manifesto. Markets are conversations, of course.

Conversations among human beings sound human. Even in 140 characters. Even more.

People recognize each other. By not selling, you are atcually making more money.

Things not to mess up(important)

•It’s a conversation thread.•Data, data, data = Metrics•Fuzzy Location #psanp

Itʼs a thread. Not one message. But a series of messages. Really important.

That being said, we have data about conversations, so thatʼs where the machines come in.

We donʼt make machines that reply to humans, we make machines that helps human reply. And find what to reply too, obviously.

Also, location is fuzzy. Itʼs ok. We use a combination of the Streaming API, precise geodata, taxonomies of expressions for places.

People Suck at naming places, but processing power sure helps to alleviate that. A bit.

Things we discoveredsurprise!

•The local voice matters•90% tweets are unanswered

Like I mentioned, the tone matters a lot @crazySFbarista is much better than @starbucks for conversations.

90% of tweets are unanswered. If you are @scobleizer with hundreds of thousands of followers or @timoreilly with (gasp) 1Milion+, maybe your questions get answered most of the times.

But for the rest of us here? Sometimes. For most new twitter users? Never.

So actually, people are ecstatic when they are found and replied. On Twitter, hashtag #FTW.On facebook it would be another story. Hashtag big brother.

Things we discoveredsurprise!

•Meaningful Conversations•Reports, Counts, Tweaks•140 = condensed meaning

Another thing we are working on, conversion. Itʼs hard to completely track sales, of course, loosely coupled.

But thereʼs lots of signals: replies, retweets, follows, clicks, checkins, trackable calls...

But even more: nice words! thank you, awesome, thanks, great, merci, loved it...

And this signal can then be parsed, analyzed, amplified.

Someone said that a tweet was a concentration of meaning. Being so short, it forces us to distill what we say. That sure rings true to me.

BUFFER SLIDE #FTWBuffer slide. Extra 20 seconds, most probably required at this point.

Hehehe.

Ok now, 10 facts about people and places.

Locked Out in London(every day)

Everyday, 5 people lock themselves out in London, UK.

Everyday. The law of large numbers is on our side now. I donʼt know if this is some early warning sing of the singularity happening, but itʼs fascinating.

So we have a locksmith category, of course. Local locksmiths can get alerts about this.

We reply for them, we have discovered 99% of small business owners are overwhlemed.A small fraction reply and tweet, but the large majority is happy not to.

(CC) zebble - http://www.flickr.com/photos/zebble/6786151/

Moving from Boston(to Palo Alto)

We did this other test in cities with Universities. Late spring and late summer are great for Movers and Storage.

So there is a certain seasonality for some services. Think snow removal (this is big where Iʼm from).

We also see other temporal hints. 11h am is, you guessed it, when people tweet about upcoming lunch.

We sometimes add event names and venues to find opportunities. Any Giants fans in the room? Yeah, tought so.

Montreal Canadiens represent a nice signal for us locally. Had to slide that reference on my Boston slide. Ok, Hockey joke, Iʼm Canadian.

http://www.google.com/search?q=I'm+ceo+bitch

Pastis in Paris(mais oui, mais où)

We partner in each new city we go in. Last year in Paris for Le Web, we worked with DisMoiOu (TellMeWhere) to find good answers.

This city is made up of arrondissements, our community managers where not that good at knowing if the 2nd intersected with the 9th (it does).

Or where to send a Québécois like me looking for a good beer. You know what they say about French beer and Belgian Wine...

If there is one thing we know, is that some local questions are easy to answer from a distance. Others arenʼt.

(CC) Tom Flemming - http://www.flickr.com/photos/tomflem/3647220170

Tooth Ache in Toledo(taking a bite at it)

Something else we learned is that cities and urban areas are sometimes under rated, but the data doesnʼt lie.

We had this request for a partnership in Toledo, Ohio. If you had asked my opinion about the opportunity volume in Toledo, I would have passed.

But I know I am ignorant so we just ask the data. We start a city for a week, harvest, index and tally up the counts.

We were blown away. Dentists in Toledo will hear about Needium any day now, if they havenʼt already.

(CC) cloud_nine - http://www.flickr.com/photos/cloud_nine/12391317/

Lingerie in Calgary(western style)

This really showed me how many preconceptions I had about the world. Even other Canadian cities (itʼs a big country, eh).

We use about 1000 expressions to match needs to categories. Explicit is easy. Who can recommend a good sushi place in SOMA?

We get the best conversations using implicit. There is this concept of “Life Events” thatʼs really helpful.

Would a cowgirl on Calgary tweet about needing new garters? Probably not.

But about an anniversary? A special date? Or just wanting to feel good? Of course. All the time.

(CC) nowhere Zen New Jersey - http://www.flickr.com/photos/98778636@N00/2225704831/

Lorem ipsum dolor adiscumName of the presenter and the time/date information

Lorem ipsum dolor sit amet, consectetuer adipiscing elit, sed diam nonummy nibh

euismod tincidunt ut laoreet dolore magna aliquam erat volutpat. Ut wisi enim ad

minim veniam, quis nostrud exerci tation ullamcorper suscipit lobortis nisl ut aliquip ex

ea commodo consequat. Duis autem vel eum iriure dolor in hendrerit in vulputate velit

esse molestie consequat, vel illum dolore eu feugiat nulla facilisis at vero eros et

accumsan et iusto odio dignissim qui blandit praesent luptatum zzril delenit augue

duis dolore te feugait nulla facilisi.

•One•Two•Three

http://needium.com

Going to Montréal(poutine, je te le dis)

Another experiment we ran last summer was with the tweet location and what we called the location intent.

We used the Twitter search API to find people in Boston talking about going to Montréal.

We then recommended hotels, a unique breakfast place and of course, some Poutine.

Trust me, it tastes better than it looks. French fries, cheese curds and gravy.

Best thing EVAR to recommend a few hours after we sent you to that Brewpub...

http://twitter.com/montrealpoutine and http://montrealpoutine.ca/

Computer Repair in Austin(expletives not deleted)

When you are runnning on a few million tweets per day, you might want to discard the noise.

We had initially setup a filter to remove swear and sex words, essentially a system wide blacklist.

Turns out that when you are trying to match for computer repairs (or a shiny new gadget),these words represent a stong positive signal of a need. We should have knwo that, intuitively, but the actual data speaks volume. I wasnʼt able to map the specifc words with the operatings systems, maybe for Ignite @ Strata ;-)

(CC) miss_rogue - http://www.flickr.com/photos/missrogue/347361369/

New York Plumbers(oh hai, I’m in deep sh*t)

Lastly, I must admit that the tough problems were not exactly where I tought they would be.

If I could, I would show you our taxonomy for Plumbing.

Material (copper, pvc, lead). Brands. Types of problems. Swear words (again). Panic words.

Also Tsunamies and Earthquakes. We had a problem with the Search API a few weeks ago, suddenly getting much less result. Dramatically.

Turns out the ops team at Twitter had to throttle down everyone because of number of requests for the events in Japan.If you had told me that an earthquake in Japan would impact my Business in Montreal... The world IS flat.

(CC) Henry Stern - http://www.flickr.com/photos/artisticrichmond/2716147621

Thanks!Sylvain CarleCTO and co-founder @ Needium

http://afroginthevalley.com/@froginthevalley

http://needium.com/@NeediumHQ

I hope you liked these anecdotes, come and see me after, I have more.

But I had to “make it quick”, as they say.

Thanks!