When firms meet in multiple markets, they can leverage punishment ability in one market to sustain collusion in another. This is the first paper to test this theory for multiproduct retailers that sell consumer goods online. With data on the universe of consumer goods sold online in Sweden, I estimate that multimarket contact increases prices. To more closely investigate what drives the effect, I employ a machine-learning method to estimate effect heterogeneity. The main finding is that multimarket contact increases prices to a higher extent if there are fewer firms participating in the contact markets, which is one of the theoretical predictions. Previous studies focus on geographical markets, where firms provide a good or service in different locations. I instead define markets as different product markets, where each market is defined by the type of good. This is the first paper to study multimarket contact and collusion with this type of market definition. The effect is stronger than in previously studied settings.