A machine selects a candidate data service to be tested for service quality. The machine performs a first-phase test of the candidate data service by providing predetermined input data to the candidate data service and measuring a first quality metric that results. The machine may determine that the candidate data service is to undergo the second-phase test based on the first quality metric, and the machine may accordingly perform the second-phase test of the candidate data service. The second-phase test may include: adding the candidate data service to a pool of five data services processing undetermined input data; routing a portion of the undetermined input data to the candidate data service; and measuring a second quality metric resultant from the candidate data service processing the routed portion. Based on the second quality metric, the machine may record whether the candidate data service is to remain in the pool.