This study assessed ultrasound scanning as a proxy for observed reproduction records, quantified the effects of lambing year and ewe age, and estimated repeatability of traits to predict current-flock gains. Data for number of lambs recorded per ewe scanned, lambs born per ewe lambed, and embryonic losses per ewe scanned were available for 2338 Dohne Merino, 1159 SA Mutton Merino (SAMM), and 138 Merino ewes on the Mariendahl experimental farm of Stellenbosch University, with 7652, 3364, and 240 ewe-year records, respectively. Merino records spanned 1990–1992, whereas the other breeds had lambing records for 1990–2016. Scan records indicated that 89.2%–95.8% of ewes scanned pregnant with multiples also lambed multiples. Embryos lost per ewe at lambing were 0.00–0.05 in all breeds. ASReml was used to fit mixed models to the Dohne and SAMM data. Lambing year and ewe age significantly affected all reproductive traits, except for ewe age effects on embryonic losses. Two-year-old ewes were more likely to be barren than their mature contemporaries, irrespective of breed. Scanning and lambing rates were highly correlated at the ewe level, suggesting that scanning is a good proxy for lambing rate in the absence of full lambing data. Age effects confirmed that an optimal flock structure contributes to a desirable reproductive output. Results indicated that embryonic losses were random and not meaningfully related to fixed or random effects. Moderate repeatability estimates for reproductive traits support low-to-moderate current-flock gains for scanning and lambing rate. Ultrasound scanning may thus be used to optimise reproduction on farms without detailed reproduction records.