The dataset is composed of the RSSI fingerprinting calibration of two malls employing four different smartphones with the random walking survey method. Moreover, the dataset includes validation routes performed with ten smartphones from various brands and labeled with the code of the shops where they were registered. For evaluation purposes, together with the samples, information related to the shopping center like the Euclidean distance between shops is provided. The files are described as follows: building.csv’: two rows with the id of mall 1 and mall 2. ‘floor.csv’: there is a row for each of the floors of the two malls. Three floors for mall 1 and two floors for mall 2 ‘zones.csv’: for each zone is indicated the id of the mall and the id floor. There are 69 zones from mall 1 and 124 for mall 2 ‘training.csv’: in this file are registered the ids and the description of the five different calibrations performed (one per smartphone) in each of the malls. ‘training_sample.csv’: each row of the file has the RSSI in dBm recorded during the off-line calibration phase together with the timestamp, channel, and MAC of the AP. ‘route.csv’: the id’s and the description of each of the validation routes are recorded in this file. ‘route_sample.csv’: with a similar structure than the training_sample.csv, the RSSI, timestamp, channel, and MAC acquired during the validation routes are stored in this file. ‘euclidea_distance.csv”: to evaluate the error of the different algorithms, this file recorded the euclidean distance from the center of a shop to each other in meters. The dataset is to be cited as follows: J. A. López-Pastor, A.J. Ruiz-Ruiz, A.J. García-Sánchez, J.L. Gómez-Tornero. Wi-Fi RSSI fingerprint dataset from two malls with validation routes in a shop-level for indoor positioning. Zenodo repository. DOI: 10.5281/zenodo.3698238