We introduce a new measure of innovation based on important product launches by public firms in the US. Our measure is based on stock-market reactions to media articles – classified by a convolutional neural network approach as referring to new product introductions – and has two distinct advantages. First, it covers the entire spectrum of industries and is not limited to products sold by retail firms. Second, we rely on collective wisdom about product value expressed through financial markets. This lends a forward-looking aspect to our measure, and helps avoid issues associated with valuing new types of output in a changing economy. Using our measure, we derive a few stylized facts. We show that product innovations are highly persistent, both at the firm- and at the industry-level. Firms that launch more new products are larger, and they typically operate in industries that are more competitive. New product introductions correlate with productivity measures at the aggregate level. However, most of these new products are launched in industries that are not among the largest employers; moreover, employment falls further following product launches.