• http://www.rimmkaufman.com/ George Michie

    Another fine post on the topic, Sid. Our system offers a wide range of models from rules based to statistical models to stats models that can accept understood weight from hold out tests for email, display, etc. However, we find the hardest and most important piece is to get the company to buy into changing the status quo, and once they’ve agreed to change, getting them to agree to a modeling approach.

    As you suggest, when the model of choice suggests shifting budgets, shifting part of the way and looking for positive impact on the P&L is the right way to go. Ultimately moving from sub-optimal to optimal should have an impact on the P&L, provided there is significant interaction, and the current model is materially worse.

    The folks who think they’ve solved this problem either don’t understand the problem or they’re lying!

  • Pat Grady

    Loved reading this, it’s the crux of all we do here.