Mentally walk the same route in order, "arrive" at each locus, see the image waiting there, and decode it back to the original item.

Recall is just the journey run in playback: start at the first locus, notice the scene you staged, translate that image back into the fact (or number, or name) it stood for, then move to the next locus. Because the route is fixed, the items come out in the exact order you stored them. If you encoded with a system (Major System, PAO), decoding is the inverse of encoding — read the image's meaning off.

Review by re-walking the palace on a spaced schedule — soon after learning, then at growing intervals.

A palace makes information easy to lay down, but like any memory it still decays without reinforcement. Walk the route again a few minutes later, then hours, then a day, then a few days — spaced repetition applied to the journey. Each successful walk strengthens the images and the links. For a one-off need (a speech tonight) a couple of reviews suffice; for durable knowledge, keep spacing the reviews out.

Tip: This is the same principle this whole app runs on — retrieval at expanding intervals.