The journey begins with “bottom-up” models in 1927, continues in psychology to 1959 and beyond. McFadden enters in the early 1970s and focuses attention on “top-down” models. Chapman and Staelin (1982) reminded us that Luce and Suppes (1965) provided direction for expanding data; however, Chapman (1984) discouraged many by suggesting that too many observations were required to be of use in real field setting. Meanwhile top-down modelers made great strides in developing ever more complex statistical models that could represent various aspects of individual differences, correlated errors and the like. Enter the DotCom boom - personalised content, one-to-one marketing, individualised search engines. All cried out for individual-level models. At the height of the boom, Sawtooth Software asked about collaborating on a way to design search engines that could incorporate tradeoffs and capture preferences – re-enter individual models. Doing this requires models of single people, or at least a way to capture their tradeoffs and preferences. Our work on Best-Worst Scaling (Finn and Louviere 1992) seemed like a potentially promising way to get more information about preferences. The question was whether it could be made to “work”. The purpose of this talk is to review the journey to the present level of capability to model individual choices from a bottom-up approach. We discuss a number of aspects related to ways to design choice experiments that can be used to model individual. We also discuss ways to analyse the data collected in these experiments, and present several illustrative examples. We conclude by discussing how far we’ve come and how much further we need to go.