We consider in this paper multi-deep automated storage/retrieval systems, where the cells capacity is strictly greater than one load. The main advantage of this class of AS/RS is a better use of space. Its main drawback is that, in order to retrieve a desired load, it is necessary to move all the loads in front of it. This is a common characteristic of all the multi-deep automated storage/retrieval systems. The number of loads to move is given by the rank of the load to retrieve. The mean value of the average access time to a given load is an important performance index for the design of such systems. This time depends on several parameters among which the average retrieval rank is the hardest to compute. It is still an open problem, even with random storage and retrieval heuristics. In this paper, our objective is to provide an empirical formula of the average retrieval rank in a multi-deep automated storage/retrieval system with random heuristics for both storage and retrieval. With this formula, it is then easy to deduce the mean retrieval time. This computation is based on multiple simulations of various AS/RS models and a regression on the obtained data. The particular case of the flow-rack automated storage/retrieval system will be considered to illustrate our contribution. It will then be possible to use this formula for other multi-deep systems.