Techniques are described herein for a beam refinement procedure between a first communications device (e.g., base station) and a second communications device (e.g., user equipment (UE)) that uses long-term feedback information from the second communications device to maintain a beam pair link established for a downlink connection. The second communications device may transmit long-term feedback information to the first communications device in response to a request or a trigger event occurring. The long-term feedback information may be sensor information measured by the second communications device or uplink training symbols. The first communications device may localize a beam search of the beam refinement procedure using the long-term feedback information. The localized beam search may be an example of a device-specific search. The first communications device may also generate a long-term map of channel conditions based on the long-term feedback data. The long-term map may also be used to localize the search.