Not every business problem involves time series, but every business has time series problems of one sort or another. Bayesian structural time series models are a flexible and powerful tool for modeling time series data. The models are additive, allowing the analyst to combine latent state components for handling trend, seasonal, regression, and other structural features. Additivity also makes it easy to place informative priors on individual components, like a sparsity-inducing spike and slab prior on a regression component when working with large numbers of contemporaneous predictors. These methods are encoded in the bsts R package (Scott, 2011), which was developed at Google to provide Bayesian time series modeling capabilities to non-experts in Bayesian modeling. The package has been used for a variety of purposes, including nowcasting economic time series, anomaly detection, forecasting, and causal inference.