The summer before my senior year, I landed an internship analyzing sales data — and discovered in week one that the company's data lived in a legacy system nobody could export from. The analyst who knew it had left, the documentation was a decade old, and my whole project depended on data I couldn't reach.
I broke it into three parallel tracks so I'd never be fully blocked: I searched the old system's user forums for export workarounds, scheduled calls with two retired employees my manager connected me to, and started manually sampling records so I'd at least have something to analyze. The forum route eventually paid off — a batch-report feature buried three menus deep could dump data as text files, and one of the retirees confirmed the field mappings.
I got the full extract with three weeks left, automated the cleanup with a script, and still delivered the analysis on time. It found that a quarter of the sales team's leads were going to a segment with the lowest close rate, which changed how they assigned territories that fall.