Clean Messy POIs with this Fun Game :)

If you’ve ever worked with point of interest (POI) data, you know how challenging it can be. It’s a hard dataset to collect and keep fresh because of its very nature: if you manage to get a comprehensive dataset, it starts to go stale immediately as places close and hours change. Many datasets have records that never existed in the first place. No company is perfect at this, and as a result there are many large datasets of varying quality.

StreetCred is creating a marketplace around this problem, and we’ve learned through our early partnerships that we can help validate and expand pre-existing datasets. If you participated in MapAustin last March, you may have noticed records we pulled in from a web crawl among other user-generated data. We learned a lot about how to crowdsource validation on incomplete records, and we’re extending those learnings to our next efforts: MapLA and MapNYC.

If you play starting June 21 (on iOS or Android), you’ll notice many incomplete records from partner datasets. Participants will earn points by creating new places as before, but also by adding more data to incomplete places or marking non-existent places as such. If you’re in NYC, you’ll also notice data from our first contest. All POI data goes stale, and we’ll be measuring carefully how our nine-month-old dataset stands up to the our validation process, which we’re calling “The Place Challenge.”

Take the Place Challenge

We believe a trustless, decentralized system is the best way to tackle the POI problem. From the beginning we’ve been honing our validation methods: if one person creates data, multiple others must validate that it exists and is correct through consensus. We’ve learned there’s a lot of value in this… some companies pay multiple dollars per record after licensing datasets just to clean it up. The best system would have validation built in, making bad data impossible.

For MapLA & MapNYC, we’ll roll out additional methods to ensure users are at the place they’re validating and also adding more valuable information through the process. First, to validate a place you’ll need to start with a new photo. Since we’ll have plenty of photos of a place before it enters validation, we have data to work with to ensure people are doing the right thing. Participants will also help check the validity of pre-existing photos:

 
002-place-challenge-step-1.png
012-place-challenge-step-6.png
 

This doesn’t extend just to photos however, we’re also applying it to other POI data attributes like hours and address:

 
009-place-challenge-step-4-ready.png
007-place-challenge-step-3-ready (1).png
 

By taking the Place Challenge, participants will collectively reveal the truth about each place. If multiple challenges indicate incorrect data, it will be flagged for others to fix. Unlike in the past, we don’t want people to focus on matching the answers of others, instead they should identify the ground truth as part of the challenge, and the StreetCred platform will take care of the rest.

We’re excited about this evolution of our validation process, which should create better data and a more enjoyable data collection game.

Randy Meech