The Humane Society states that over 3 million dogs enter animal shelters around the United States each year, and around 2 million dogs are adopted each year. Shelters are understandably busy, noisy, and fast-moving places where many challenges present themselves. The ability to correctly anticipate how the coming days, weeks, and months will go would allow shelter managers to allocate resources more effectively. Our group sought to leverage machine learning tools and 100,000s of observations over the last decade to predict animal shelter intakes, outcomes, and adoptions. We developed time series models which include macro-level features and can predict the number of intakes and outcomes per day, week, and month with over 90% accuracy. Additionally, we achieved over 70% accuracy exploring how random forest can be used to get a paw up on predicting adoption rates with shelter-level features.