Drew leads the FinOps sourcing team for Hulu with Anthony Logan focusing on the technical aspects of cloud. At the beginning his team had no mandate from the top, which allowed them to focus on policies. FP&A was lacking technical depth while engineers were not concerned with cloud cost. Hulu is primarily using AWS with some workloads in data centers. Drew’s team ultimately reports to the CFO.
Forecasting reached a turning point when engineers asked for 4.5x of the previous year’s spend for the next budget cycle. The decentralized engineering teams made pricing decisions on products without understanding the underlying costs. The company’s objectives and key results (OKRs) put the viewer first. While Drew was able to connect spend to revenue, he was concerned about over spending.
A key OKR around accuracy to forecast was added for all engineers. Drew’s team started with good data and made sure they can trust the numbers. By driving visibility he was able to build demand-driven forecasting, which in turn allowed them to optimize spend. By identifying the top spending teams and learning about their workloads, his team were able to get the majority of the forecasts in line with expectations.
Drew is using a hybrid forecasting model, where he uses business drivers for top services and trend based forecasting for the long-tail. This allows him to update forecasts every month with an average variance of about 2.5% and the worst case being about 9% variance. To get there, Drew decided to forecast by workloads instead of cloud stock keeping units (SKUs). However engineering volatility is challenging, for example predicting load testing or new services. For new services Drew uses swag estimates to track these special projects on their roadmap.
Our learnings are the following: A tagging strategy is imperative. Partnering closely with the highest cost workloads works very well. Start building relationships with CTO and CFO, then engage with the engineering leads.