In this case, bad measurements are OK. They average out over time. If the bias is one way, then one type of agent (seller or buyer) gets a bit more pay than the other, but this is not a disaster and is compensated by agents competing to work on the higher-paying side of the transactions.