The following are my notes from the presentation given by Simon Callan (Foursquare) at MRA’s 2016 Corporate Researcher Conference. Due to limited editing there will be typos.
Original premise of FourSquare – use your mobile phone to find cool places and compete with your friends. Still has two consumer apps to use on your phone.
Very much now on personalized city guides. After a week of using FourSquare it will start pinging you and suggest places to go to.
Swarm — is a game, allows you to check in and compete with friends over who checks in most at a certain place.
FSQ: These apps generate a lot of data – 500 mllion photos, 87 million tips etc. From these can start mining tastes.
Swarm: 10 billon+ global check-ins, 8 million check-ins/day, 85 million public places and 105 million total places.
What do they do with it?
FSQ powers the geolocation of Waze, Uber, Twitter and What’s App.
Last year made a prediction of how many new iPhones were sold based on foot traffic to store. Predicted 13-15 million, and according to Tech Insider predictions “right on the nose” – closer than the analysts.
A lot of hedge and quant funds for stock predictions. Retail and CPG use it for customer analysis.
Predicted Chipotle’s sales dropped 30%; actually dropped 29.7%.
Location is hard to do, which is why it has taken FSQ several years to do.
FSQ can tell with confidence when someone is in an Apple Store, and stays there for a few minutes. Not looking at users being near places, need to be in the location.
Q: How do they turn the signals into data?
Have 55 million active monthly users global (25% in US). Look at this base and carve out a panel of about 12 million that they link to census data. Generates a total of 300 million visits a month.
Apply normalization to panel, weight census to gender for example. Also have to account for changes in app that might impact visit counts.
What does Foursquare provide:
- Store level data – anyone can buy every single store – foot traffic, demographics, where they go before and after and tastes
- Chain level data – trends across chains
- Market share reports
Turning this into insights
- Can use to look at competitive reports of foot traffic counts between different stores within the same category
- Can do this at a city level, to see which are under-performing or over-performing
- Can see what people visit after they visit a certain type of store
- No need for expensive beacons, surveys
- Is highly granular
- Is near real-time
- Provides context to other data (transactions, staffing, locations, media)