Evaluation of the Aeroplan distinction status program desirability

点击翻译按钮获取中文摘要

The Aeroplan Distinction Status program was launched in 2014 by our partner Aimia Canada Inc. to give**further rewards and recognition to the program's most engaged and active members. Under that program,**members qualify for one of three tiers based on eligible mileage accumulation from the previous calendar year.**Each tier offers benefits such as rewards and access to promotions. Our partner wants to determine if the status**program has an influence on members earn behaviour, and if so, for what type of members. The difficulty in**this project is to detect specific changes in a time series that has a natural variability that may include**non-stationary components, trends and seasonal components. The detection of Earning Rate (ER) local changes**that are influenced by status can be regarded as a two-step procedure. First, one must detect local variability**changes in the ER time series. Second, on must show that the times of such changes coincide (i.e. are close to)**with the critical times associated with status qualification thresholds. To establish that such coincidences are**not accidental, they should be observed repeatedly for several critical times. Obtaining statistically significant**results may require the grouping of several members having similar ER behaviours. The project has two**objectives:**1) Detect local changes in the ER local variability via local statistical indicators and correlate them with**qualification threshold times. 2) Detect local changes in the ER local variability via deviations from best linear**prediction and correlate them with qualification threshold times.**We will achieve objective 1) by searching for coincidences between local statistical measures (to be**developed) and critical times associated with status qualification thresholds. We will achieve objective 2) by**first developing a linear predictor of the ER series, and then by examining the prediction deviation (prediction**minus actual value) for each ER series. This deviation should reveal changes in the ER variability that could be**indicative of behaviour change.