Sicilo-Sarde dairy flock test-day data were analysed (1) to check predicting ability of ten milk recording procedures, based on information from a single monthly or bimonthly milking and (2) to look for the best pair of factors for predicting daily milk yield from either morning or afternoon milkings for the alternate testing method (AT). Prior to analyses, data were adjusted for variation factors with known significant effect. Number of test-day observations was 4755 from 458 lactations of 288 ewes. Losses of precision associated with simplified methods were evaluated by comparing estimated lactation yields with those observed both in a reference plan, where the two daily milkings were recorded at 2-weekly intervals and in the official A4 milk recording (monthly recording of the two daily milkings). Ranking of simplified plans based on their accuracy was still the same under either monthly or twice-monthly frequency. Estimates of lactation yields showed more lack of precision when the usual monthly designs were compared with twice-monthly sampling of both a.m. and p.m. milkings. Losses of precision ranged from 5·1%, for the A4 plan, to 14·6%. Thus, milk production should be estimated from both mikings when their measurement is economically feasible. When only one milking is weighed, designs based on—or beginning with—the a.m. milking were more accurate. There was no evidence of improvement in sampling accuracy by adjustment for milking interval or production level. For practical and economic reasons, the design alternating a.m. and p.m. milkings, without adjustment, is suggested for ovine milk recording. In such a case, new factors for estimating daily production from the measured milking (1·7, 2·3) should be used rather than the factors currently in use (2, 2) because they improved the loss of precision by more than 13% in a monthly recording system.