Background Increasing insecticide costs and constrained malaria budgets may make universal vector control strategies unsustainable in low transmission settings. We investigated a reactive, targeted indoor residual spraying (IRS) strategy. Methods A cluster-randomised, open-label, non-inferiority trial (ClinicalTrials.gov: NCT02556242) compared reactive, targeted IRS (TIRS) to current standard IRS (SIRS) practice in South Africa over two malaria seasons (2015–2017). In SIRS clusters, programme managers conducted annual mass spray campaigns prioritising areas using historical data, expert opinion, and other factors. In TIRS clusters, only houses of index cases (identified through passive surveillance) and their immediate neighbours were sprayed. The non-inferiority margin was 1 case per 1000 person-years (py). Health service costs of ‘real-world’ implementation were modelled from primary and secondary data. Incremental costs per disability-adjusted life-year (DALY) were estimated and deterministic and probabilistic sensitivity analyses conducted. Findings Malaria incidence was 0·95 (95%CI 0·58–1·32) and 1·05 (95%CI 0·72–1·38) per 1000 py in the SIRS and TIRS arms, respectively, corresponding to a rate difference of 0·10 (95% CI -0·38–0·59) per 1000 py, demonstrating non-inferiority of TIRS (p<0·001). Per additional DALY incurred, TIRS saved $7,845 (95% CI $2,902–$64,907), giving a 94-98% probability that switching to TIRS would be cost-effective relative to plausible cost-effectiveness thresholds for South Africa ($2,637-$3,557 per DALY averted). Depending on the threshold used, TIRS would remain cost-effective at incidences <2·0–2·7 per 1000 py. Findings were robust to plausible variation in other parameters. Interpretation TIRS was non-inferior, safe, less costly, and cost-effective compared to SIRS in this very low transmission setting. Saved resources could be reallocated to other malaria control and elimination activities. Funding Joint Global Health Trials.