Backend Engine Efficiency

What approach would you take to assess the comparative efficiency of two backend engines used for the automated creation of Meta/Facebook 'Friend' recommendations?