AutoML refers to Automated Machine Learning. Machine Learning (ML) algorithms build models largely automatically once they get cleaned up data, but still, there are quite a bit of manual work to be done to make ML work, for example, data preprocessing, feature selection/generation, and hyperparameter tuning. AutoML aims to reduce these manual work so that human experts can focus on more sophisticated analysis and speed up the development process.
We have an in-house AutoML package which we use for benchmarking purposes. It provides a useful baseline model upon which to improve and customize. We used the AutoML package for price variance comparison as it is a good way of making comparable baseline models for each client in a vertical.
In the above chart, we see a high variance in price variances among clients in the same vertical. This may be due to different levels of commercial control over pricing and governance. With this baseline result as a starting point, MarketRedesign’s customized pricing model can further identify pockets of price potentials more accurately and help improve the client’s value recovery.