I ask the question most forecasting projects skip: is this data even stable enough to model? Using five years of Apple, Microsoft, and Alphabet share prices, I test for stationarity with the Augmented Dickey-Fuller test, compare raw prices against daily log returns, and run structural-break analysis to show how major market events shift underlying relationships. The result is a practical case for why financial models should start with understanding data behavior — not with prediction. Built in Python using real Yahoo Finance data.