We show that the random insertion method for the traveling salesman problem (TSP) may produce a tour Ω(log log n/log log log n) times longer than the optimal tour. The lower bound holds even in the Euclidean Plane. This is in contrast to the fact that the random insertion method performs extremely well in practice. In passing, we show that other insertion methods may produce tours Ω(log n/log log n) times longer than the optimal one. No non-constant lower bounds were previously known.