The author provides a simple approach to the problem of linking attribution effects over time:  Just Don't Do It.  Instead he suggests a "long single-period" approach using style-based benchmarks and simple subtraction.