Introduction: Evidence about mobile health (mHealth) approaches to manage diabetes shows modest effects on outcomes, but little is known about implementation variability. This is a process evaluation of an mHealth intervention to improve diabetes self-management through Short Message Service (SMS) provision in three diabetes care programmes in the Democratic Republic of Congo (DRC), Cambodia and the Philippines. Methods: The intervention involved Diabetes Self-Management Support via text messages. The content and process of the intervention is based upon the core principles of diabetes self-management and behaviour theory. In each country, messages were sent by project managers to 240 participants in each country, who were randomly assigned to the intervention group. Contracts were negotiated with national phone providers and open access software was used to send the messages. Participants received a mobile phone and SIM card. We analysed data about the implementation process over a one year period. Results: The mean monthly number of messages delivered to recipients’ phones was 67.7% of the planned number in DRC, 92.3% in Cambodia and 83.9% in the Philippines. A telephone check revealed problems with one-third of the phones, including breakage, loss and cancelled subscriptions. The number of people reached at least once was 177 (70.0%) in DRC; 147 (60.7%) in Cambodia; five in the Philippines (2.0%). Those reached each time was 144 in DRC (56.9%), 28 (9.9%) in Cambodia, none in the Philippines. People used their phone more frequently than before the intervention. Discussion: Implementation of the intervention meets constraints at every step in the process. Barriers relate to the technology, the context and the participants.