In 2013, a small team of technologists set out to unite the world by connecting them to the people and places around them. It sounded easy enough ... right?
And so began SocialRadar: a networking app, which leveraged user locations and mapping data to give you context about your surroundings.
Step 1: Master user location.
All of the documentation we read on GPS/CoreLocation explained: "up to 3-meter accuracy!" Perfect! That would fit our needs. But, as it turns out, GPS is a very unreliable signal and as a developer, you have very limited access to it.
Thus, we had to refine that signal. We combined it with other data streams, came up with predictive logic and minimized battery loss.
In the end, we got pretty good at user location. Our accuracy was like none other, and we were pretty darn proud of ourselves.
Step 2: Mapping.
Just because you have an accurate latitude and longitude coordinate for a user, doesn't mean you know what store or restaurant your user is standing in. It means you have a latitude and longitude coordinate – two abstract numbers devoid of any practical meaning. Ugh!
It turns out, we needed to take that coordinate and reverse geocode it. Or, in layman terms, we wanted to match that point to a business location. Again, that seemed easy enough. We know where the businesses are, don't we?
The logical next step, turn to a major third party provider in the mapping industry. They have to know where points of interest (POIs) are, right? That's their job. So we sent them a user's latitude and longitude coordinate to reverse geocode, and what we got back was a list of possible businesses. They said our user was standing close to several locations, but they didn't know exactly which one. Wait, what? ¯\_(ツ)_/¯
So, we looked into the issue.
As it turns out
Major third party providers assign one latitude and longitude coordinate to every business so every business is represented by one point on a map. And as we all know, a single point is not representative of the physical location of a business. Businesses have a shape to them with clear boundaries.
We looked into alternatives and they all had the same thing, everyone in the industry had the same data. And, worst of all - most of their points weren't even accurate at all! A third of the points were in the middle of the street – not even in a building entirely.
What we needed
Third party providers didn't have the data we wanted or the quality we needed. What we were looking for was:
- A way to see the entire geometry of a business. What are the stores boundaries?
- Exactly where the front door is. How does one enter this business?
- Accurate locations. All of the venues should be in a building and the right one.
The Only Option
We had to build our own POI database. One that is independent of any other dataset and that is human verified. And so, that's what we did.
Currently, SocialRadar POIs have 12 different location attributes, including a precise bounding box for over 12M POIs in the US.
And, as it turns out: a lot of people have been looking for this level of accuracy too.
Stay tuned for blog #2: How in the He** Did We Do That?