We think backfill bias is a natural result of the research process and arises when a manager is constructing an entirely new program or adding a new model to an existing program. In its simplest form, a manager begins by developing an investment thesis, then constructs the strategy using in sample data and tests on an out of sample data set.