The subject technology receives, from a first client device, a set of requests for recording user activity detected on a mobile application, each request includes information corresponding to a set of events based on the user activity that has occurred on the mobile application. The subject technology determines that a first request from the set of requests includes an image and a hash of the image. The subject technology stores the image and the hash of the image in a cache. The subject technology determines that a second request from the set of requests includes the hash of the image, the second request being different than the first request. The subject technology retrieves, based on the hash of the image, the image from the cache. The subject technology generates a markup language element based at least in part on the retrieved image from the cache.