This volume is largely about nontraditional data; this paper is about a nontraditional visualization: classification trees. Using trees with data will be new to many students, so rather than beginning with a computer algorithm that produces optimal trees, we suggest that students first construct their own trees, one node at a time, to explore how they work, and how well. This build-it-yourself process is more transparent than using algorithms such as CART; we believe it will help students not only understand the fundamentals of trees, but also better understand tree-building algorithms when they do encounter them. And because classification is an important task in machine learning, a good foundation in trees can prepare students to better understand that emerging and important field. We also describe a free online tool--Arbor--that students can use to do this, and note some implications for instruction.