What support does research software need? Early lessons from the Research Software Maintenance Fund

What support does research software need? Early lessons from the Research Software Maintenance Fund Neil Chue Hong, Anna Roubíčková, Catherine Inglis, Giacomo Peru, Kathleen Glass, Selina Aragon University of Edinburgh Research Software Maintenance Fund £4.8m fund run by the Software Sustainability Institute on behalf of UKRI Aims: ●Understand how to better fund software maintenance, including what types of activity should be prioritised for funding; ●Support key software used by the community; and ●Diversify the people and software supported to encourage a wider range of contributions. Two rounds of funding, the first encompassing large awards (up to £500k for two years) and small awards (up to £150k for one year) Two stage process: expressions of interest and full applications An extremely high level of interest Highly-attended webinar, hundreds of questions. Geo-political uncertainties led to a much higher than anticipated international involvement in submissions. 360 Expressions of Interest were submitted to the call, with the large majority being in-scope - we anticipate being able to fund about 5 large and 10 small awards in this round. This meant that we had to change our previous aim of allowing all in-scope EoIs to submit a full application ●Considered restricting number of applications per Lead (would not have made significant difference) or per Institution (felt to be unfair at this stage) ●Ideal funding ratio would be 20%, but this would only take 75 applications through, so aimed to take 150 through ●Submitted: 172 large awards. 188 small awards ●Accepted: 69 large awards, 81 small awards Extremely high fit to call 126 applications scored top marks across all key criteria, a further 122 only dropped a single grade in one key criterion Only seven EoI could be considered out of scope, and even then only one was clearly out of scope Types of activity requested Almost all the EoI’s included some form of technical activities. Large awards were more likely to include Community and Governance activities than small awards The highest scoring EoI’s for large awards were more likely to include Governance activities (42%) The lowest scoring EoI’s for large awards were least likely to include community activities (77%) The highest scoring EoI’s for small awards were more likely to include Documentation (88%) The lowest scoring EoIs were more likely to include training activities (71%) Total Large Small Technical 99% 99% 99% Community 81% 87% 75% Documentation 84% 88% 81% Training 65% 69% 62% Governance 27% 35% 20% Career Stage Mid-career leads were more likely to be highly scored. EoI leads were overwhelmingly White British or other White background. No significant difference between the three pools with regards to whether leads indicated that they were disabled or not. Large % > 700 Large % 670-700 Large % < 670 Small % > 700 Small % 670-700 Small % < 670 Early 26.7% 31.3% 24.1% 28.9% 34.4% 37.5% Mid 41.9% 37.5% 37.9% 50.0% 40.6% 33.3% Established 29.7% 29.7% 36.2% 18.4% 21.9% 25.0% Not specified 1.8% 1.6% 1,7% 2.6% 3.1% 4.2% Large (total EoIs) Large > 700 Large 670-700 Large < 670 Small (total EoIs) Small > 700 Small 670-700 Small < 670 Physical Sciences 77 40.6% 44.8% 50.0% 97 59.2% 48.4% 43.8% Engineering & Technology 55 28.1% 32.8% 36.0% 51 28.9% 23.4% 29.2% Computer Science and Mathematical Science 91 42.2% 58.6% 60.0% 88 59.2% 37.5% 39.6% Life Sciences 76 57.8% 32.8% 40.0% 74 32.9% 53.1% 31.3% Social Sciences 31 10.9% 15.5% 30.0% 31 5.3% 28.1% 18.8% Arts and Humanities 15 6.3% 5.2% 16.0% 15 3.9% 15.6% 4.2% Other 23 12.5% 15.5% 12.0% 19 7.9% 10.9% 12.5% Research areas Life sciences more highly represented in high scoring large awards; Physical, Computer and Mathematical Sciences more highly represented for small awards Social sciences and arts and humanities show drop off and are underrepresented Benefit criteria Some EoI’s demonstrated a large international user base but failed to clearly demonstrate use in the UK In general, the landscape analysis was very good Activities criteria There were EoIs which had a significant amount of novel development, that scored lower In some, the alignment between objectives, activities and impact was vague, even for an EoI The community feedback criterion had the most scoring variance. ●In lower scoring applications, community communications mechanisms were only one way, e.g. a broadcast mailing list or documentation. ●The highest scoring applications offered ways for users to raise issues or feature requests, or even to create or improve ways in which the community can contribute to the software development roadmap.