Techniques are disclosed for caching provenance information. For example, in an information system comprising a first computing device requesting provenance data from at least a second computing device, a method for improving the delivery of provenance data to the first computing device, comprises the following steps. At least one cache is maintained for storing provenance data which the first computing device can access with less overhead than accessing the second computing device. Aggregated provenance data is produced from input provenance data. A decision whether or not to cache input provenance data is made based on a likelihood of the input provenance data being used to produce aggregated provenance data. By way of example, the first computing device may comprise a client and the second computing device may comprise a server.