The “freemium” model for digital goods involves selling a base version of the product for free, and making premium product features available to users only on payment. The success of the model is predicated on the ability to profitably convert free users to paying ones. Price promotions (or “sales”) are often used in freemium to induce the conversion. However, the causal effect of exposing consumers to such inter-temporal price variation is unclear. While sales can generate beneficial short-run conversion, they may be harmful in the long-run if consumers inter-temporally substitute purchases to periods with low prices, or use them as signals of low product quality. These long-run concerns may be accentuated in freemium, where the base version is sold for free, so that sales form extreme price cuts on the overall product combination. We work with the seller of a free-to-play video game to randomize entering cohorts of users into treatment and control conditions in which promotions for in-game purchases are turned on or off. We observe complete user behavior for half a year, including purchases and consumption of in-game goods, which, in contrast to much of the extant literature, enables us to assess possible substitution over time in consumption directly. We find that conversion and revenue improve in the treatment group; and detect no evidence of harmful inter-temporal substitution or negative inferences about quality from exposure to price variation, suggesting that promotions are profitable. We conjecture that the zero price of the base product that makes its consumption virtually costless, combined with the complementarity between the base product and premium features can help explain this. To the extent that this holds across freemium contexts, the positive effects of promotions documented here may hold more generally.