In transfer pricing arm’s length prices are often determined based on benchmark studies. These studies produces a range of figures. In some cases, not all comparable transactions will have a relatively equal degree of comparability. Where every effort has been made to exclude points that have a lesser degree of comparability, it may still be the case, that what is arrived at is a range of figures for which comparability defects remain that cannot be identified and/or quantified.
Substantial deviation among points in that range may indicate that the data used in establishing some of the points may not be as reliable as the data used to establish the other points in the range.
In such cases, if the range includes a sizeable number of observations, statistical tools (e.g. the interquartile range or other percentiles) can enhance the reliability of the analysis. It may also be appropriate to use measures of central tendency (for instance the median, the mean or weighted averages) to determine the arm’s length price applied, in order to minimise the effect of unknown or unquantifiable remaining comparability defects.
See on this issue, TPG 2017, para 3.55 – 3.62