Lessons from Everyblock

Adrian Holovaty, founder of ChicagoCrime.org and Everyblock.com, spoke at O'Reilly Media's Where2.0 conference. Video of the entire event is at Blip.tv. Here's a breakdown of what Holovaty says we can learn from Everyblock:

1: Take advantage of existing data

Plenty of sites start out by asking for contributions. Everyblock doesn't. Its first mission is to make data that's already floating around the internet and locked in government file cabinets available and easy to access. Tips for getting data:

• Be nice: People will help you out if you're polite. Duh.
• Governments should focus on services: Making mashups and analyzing data is our job. The government's job is to solve problems and serve its citizens.
• Plot cities/agencies against each other: Tell New York what Chicago is providing. Let one agency know that their bureaucratic rivals are way more open and cooperative. It really does work.

2: The more local it gets, the more effort it takes; but the reward is bigger

Every city has different ways of managing data, and there are no national databases for things like building permits, restaurant inspections or even crime (San Francisco, for example, only specifies a zip code).

Private sites like Flickr, Yelp and others have APIs that work anywhere. A photo in San Francisco is the same as a photo in Chicago.

But it's worth doing the work, Holovaty says. No one else is aggregating this much information on such a focused area.

3: Embrace hypertext: "Be of the web, not just on the web"

One question to ask when building a site like Everyblock or Chicago Crime: "Will my site work without maps?"

Individual data points--a crime, a liquor license, a neighborhood--all have permanent URLs. All can be found by humans and by Google.

All this makes information easy to find, and easy to share.

4: Move beyond points

Some news applies to a whole neighborhood. Sometimes to a street. Sometimes a block.

When news affects an area, show an area, not a point. This is a bit trickier, technically, but data is more meaningful when it's applied the right way. If information is only as specific as a city, your map doesn't need to zoom down to street level. Keep it where it's relevant.

5: Roll your own maps

Google maps are designed for getting directions. They include one-way streets, bus stops and BART stations. Plus, Google (or Yahoo, Microsoft, etc) decides how they look.

As Holovaty says: You wouldn't make your corporate website with some generic Wordpress template. Why do that with your maps? Here's how Everyblock made its maps.

Also notable: Everyblock takes out names to protect privacy. "We don't want to make people Google-able."

On how Everyblock will make money: "We have no freakin' idea. We might go for another grant. We might go VC. We might magically dream up a business model."

The last question is especially noteworthy: How do you get boundary data?

"We've had good luck with governments," Holovaty said. And that's a bigger thing than most people realize. American locations geocode pretty well. Try making a Google map in China, and you'll find yourself staring at a sparsely populated satellite image with no streets and little searchable beyond major cites. And that's an improvement.

When I first tried making a Dalian map, I had to hunt for landmarks that could be seen from space. Searching "China" returned no results. We essentially did the same thing for the Dalian map (all our bar, cafe, restaurant and event listings) at DalianDalian.

It is possible to get street-level data in China, but only if you read Chinese. Go to ditu.google.com and you'll see what I mean. Here's what Xi'an looks like in Ditu, and here it is in English.

Before: The Pottery Barn theory of blogging | After: Making multimedia a habit