The aim of this study was to identify risk factors for culling, sales and deaths in intensively managed dairy goat herds in New Zealand. A data set provided by the New Zealand Dairy Goat Cooperative (n = 13,197 does) was analyzed using a Cox proportional hazard model. The outcome of interest was length of productive life (LPL), defined as the number of days from the date of second kidding to the date of removal from the herd or the date on which follow-up was terminated, whichever occurred first. Milk solids yield in the first lactation (MSL1) as a predictor of LPL was parameterized in the model as a penalized spline term. To account for MSL1 violating the proportional hazards assumption of the Cox model, LPL was divided into two intervals: T1 (less than or equal to 730 days from the date of second kidding) and T2 (greater than 730 days from the date of second kidding). MSL1 was then included in the model as a time-dependent covariate. A frailty term was included in the model to account for unmeasured, herd-level effects on LPL. During T1, the daily hazard of removal for does that produced 80 kg milk solids in the first lactation was 0.84 (95% CI 0.58–1.23) times the daily hazard of removal for does that produced 30 kg milk solids in the first lactation. During T2, the daily hazard of removal for does that produced 80 kg milk solids in the first lactation was 1.44 (95% CI 0.79–2.65) times the daily hazard of removal for does that produced 30 kg milk solids in the first lactation. We conclude that involuntary losses may be avoided if high MSL1 yielding does are preferentially managed from 2 years beyond the date of second kidding.