PHD decided to shift its perspective – rather than media planning, the agency wanted tools that would help Honda predict the probability of being able to hit its sales targets. So, the agency reinvented its approach to marketing mix modeling to get more precise answers, faster and more regularly. It introduced an application that automates data extraction. It also re-wrote its modelling code – adapting it so that it could be integrated into an automated workflow. Then, it hired a statistician and used machine learning to better predict sales for Honda vehicles. Now, the agency can predict results within 3% of accuracy. In the first quarter, the new modelling system informed decision-making on pricing, budgeting, media mix and flighting for key vehicles. As a result, sales for Honda models outperformed competitors by 10% nationally and 13% regionally.