The initial version uses the Google Prediction API as the machine learning backend, and obviously some tweaking would need to be done to the model to get it to 99% right, but I'm confident that for verticals like event tickets this service would be very useful. Imagine Coachella tickets not selling out within 5 minutes or a play at a local theater having an automatic discount the night before the show if half the seats are still unfilled.
{"item": 150242, "price": 1999.00, at: "2014-08-03T12:40:34Z"}
{"item" 150242, "quantity_left", 48, "surge_price": 2033.00, "original_price": 1999.00, "percent_change": 1.7 }We also have implemented a "real-time" aspect whereby the price can also fluctuate based on the number of users currently viewing/purchasing an item vs. the average number of users usually viewing the item (for example when Outside Lands begins selling their tickets at 10am PST). Other factors taken into consideration (as of now, but more to come) are: inventory_remaining, and profit_margin (i.e. how much you make on each sale).