The return of samples from known locations on Mars is among the highest priority goals of the international planetary science community. A possible scenario for Mars Sample Return (MSR) is a series of 3 missions: sample cache, fetch, and retrieval. The NASA Mars 2020 mission represents the first cache mission and was the focus of the CanMars analogue mission described in this paper. The major objectives for CanMars included comparing the accuracy of selecting samples remotely using rover data versus a traditional human field party, testing the efficiency of remote science operations with periodic pre-planned strategic observations (Strategic Traverse Days), assessing the utility of realistic autonomous science capabilities to the remote science team, and investigating the factors that affect the quality of sample selection decision-making in light of returned sample analysis.CanMars was conducted over two weeks in November 2015 and continued over three weeks in October and November 2016 at an analogue site near Hanksville, Utah, USA, that was unknown to the Mission Control Team located at the University of Western Ontario (Western) in London, Ontario, Canada. This operations architecture for CanMars was based on the Phoenix and Mars Exploration Rover missions together with previous analogue missions led by Western with the Mission Control Team being divided into Planning and Science sub-teams. In advance of the 2015 operations, the Science Team used satellite data, chosen to mimic datasets available from Mars-orbiting instruments, to produce a predictive geological map for the landing ellipse and a set of hypotheses for the geology and astrobiological potential of the landing site. The site was proposed to consist of a series of weakly cemented multi-coloured sedimentary rocks comprising carbonates, sulfates, and clays, and sinuous ridges with a resistant capping unit, interpreted as inverted paleochannels.Both the 2015 CanMars mission, which achieved 11 sols of operations, and the first part of the 2016 mission (sols 12–21), were conducted with the Mars Exploration Science Rover (MESR) and a series of integrated and hand-held instruments designed to mimic the payload of the Mars 2020 rover. Part 2 of the 2016 campaign (sols 22–39) was implemented without the MESR rover and was conducted exclusively by the field team as a Fast Motion Field Test (FMFT) with hand-carried instruments and with the equivalent of three sols of operations being executed in a single actual day. A total of 8 samples were cached during the 39 sols from which the Science Team prioritized 3 for “return to Earth”. Various science autonomy capabilities, based on flight-proven or near-future techniques intended for actual rover missions, were tested throughout the 2016 CanMars activities, with autonomous geological classification and targeting and autonomous pointing refinement being used extensively during the FMFT. Blind targeting, contingency sequencing, and conditional sequencing were also employed.Validation of the CanMars cache mission was achieved through various methods and approaches. The use of dedicated documentarians in mission control provided a detailed record of how and why decisions were made. Multiple separate field validation exercises employing humans using traditional geological techniques were carried out. All 8 of the selected samples plus a range of samples from the landing site region, collected out-of-simulation, have been analysed using a range of laboratory analytical techniques. A variety of lessons learned for both future analogue missions and planetary exploration missions are provided, including: dynamic collaboration between the science and planning teams as being key for mission success; the more frequent use of spectrometers and micro-imagers having remote capabilities rather than contact instruments; the utility of strategic traverse days to provide additional time for scientific discussion and meaningful interpretation of the data; the benefit of walkabout traverse strategies along with multi-sol plans with complex decisions trees to acquire a large amount of contextual data; and the availability of autonomous geological targeting, which enabled complex multi-sol plans gathering large suites of geological and geochemical survey data.Finally, the CanMars MSR activity demonstrated the utility of analogue missions in providing opportunities to engage and educate children and the public, by providing tangible hands-on linkages between current robotic missions and future human space missions. Public education and outreach was a priority for CanMars and a dedicated lead coordinated a strong presence on social media (primarily Twitter and Facebook), articles in local, regional, and national news networks, and interaction with the local community in London, Ontario. A further core objective of CanMars was to provide valuable learning opportunities to students and post-doctoral fellows in preparation for future planetary exploration missions. A learning goals survey conducted at the end of the 2016 activities had 90% of participants “somewhat agreeing” or “strongly agreeing” that participation in the mission has helped them to increase their understanding of the four learning outcomes.