Neural networks often feel like black boxes. In this article, we build a neural network regressor from scratch using only Excel formulas. By making every step explicit, from forward propagation to backpropagation, we show how a neural network learns to approximate non-linear functions with just a handful of parameters.