We analyze the energy consumption data of tenants in office buildings in Tokyo and show that it is able to estimate the specific energy consumption of each tenant with a high degree of accuracy by using a linear mixed model. We obtain the result that attributes which strongly influence on the basic unit are temperature, tenant type and so on. From the observation of random effects, we know that the main factor to the basic unit is the individuality of tenants. We also clarify effective power saving behaviors after the Great East Japan Earthquake by a random effect.