Policymakers can choose from a variety of fiscal stimulus levers: conditional transfers, unconditional transfers, or direct purchases from certain industries. This choice is complicated by the rich network structures that connect households and industries via employment, directed consumption, and input-output linkages. We study this problem in a model with household heterogeneity in MPCs, directed consumption patterns, and exposure to industry and regional shocks. Theoretically, we express fiscal multipliers in terms of estimable sufficient statistics, and we decompose them into three network effects on top of a standard Keynesian multiplier. Empirically, we find that targeting fiscal policy is important, but simple. First, optimally targeting fiscal stimulus generates twice as much amplification in GDP as untargeted policy. Second, owing to the empirical absence of two of the three network effects, a simple fiscal policy that targets households based on their MPCs is close to maximally expansionary and optimal.