This paper uses an experimental dataset under development at Statistics New Zealand, the Linked Employer-Employee Database (LEED), to examine labour market outcomes for individuals previously receiving welfare benefits. This data contains monthly information on individuals’ benefit and earnings receipts over the three-year period from April 1999 to March 2002. In contrast to most administrative data used for benefit transition analyses, which can only identify benefit versus non-benefit states, the data used in this paper reveals whether or not an individual is employed both when they are on and off benefits. We exploit these features of the data to examine the labour market outcomes for the working-age population who ever receive benefits over the sample period, and focus on two broad issues. First, we provide a critical assessment of the adequacy of using off-benefit at a future point in time as a measure of a successful benefit-to-work transition, and compare it with the alternative in-employment measure. We find that off benefit rates are typically 85-90 percent over the 18 months following a reference benefit-spell, compared to in-employment rates of only 55-60 percent. Also, for those who are off-benefit, women, older individuals and those living in Auckland, are less likely to be employed. Second, we examine the relationship between an individual’s benefit-spell duration and their employment experience, both during their benefit-spell and prior to the spell, on their post-benefit labour market outcomes. We find that longer benefit-spell duration has a negative effect, and previous employment experiences have positive effects on subsequent outcomes.