I would recommend opening a support case so it can be properly looked into.
If they create a forecast via weighted-historical index method, then the quality of the forecast relies on the quality of their historical day selections and weighting. If they are not using automatic best method forecast generation, I would first suggest using that method. They could even create a forecast for a period in the past and check actuals versus forecast, forecast error, and forecast accuracy in the forecast viewer/editor.
AHT and offered variances are specifically due to historical trends, patterns, etc. If those differ wildly either because historically they did and/or because of changing configuration or assumptions week-to-week, then obviously predicted service performance (e.g., SL, ASA, ABD) and staffing requirements will differ significantly.
------------------------------
Jay Langsford
VP, R&D
------------------------------