In one embodiment, scanning data is received for wireless access points whose wireless signals were observed by the one or more scanning devices in a target area. For each group of scanning data that shares a common identifier of a Wi-Fi access point, a centroid of the scanning data is determined, a set of scanning data of the group that exceeds a threshold distance from the centroid is designated as having potential error, and a location of the Wi-Fi access point is calculated. The calculated location of the Wi-Fi access point for each group of scanning data is stored in a reference database. Upon receiving a request for a location of a mobile device, an identity of one or more Wi-Fi access points in range of the mobile device is determined, and the location of the mobile device is computed using calculated locations from the reference database.