Censuses are used by governments to see what is going on with people. They collect a huge amount of data that society needs in order to describe the present and plan for the future. However, censuses have a very short useful life as the population is always changing. Surveys are used to get updates on the population in the period before the next census, or if the information wasn’t taken in the last census. Imagine if a survey is carried out on the whole population every time that information is needed. It would soon get very costly. But what if a much smaller group (or sample) of people could tell us what we need to know about the whole country? Some sections of the country will be more costly to survey than others so we need to bear this in mind when choosing our sample. Mervyn has developed software that selects a minimum number of people necessary to give an accurate picture of what everyone is doing across the country while keeping the costs down. His software groups the population by common characteristics and searches for the best grouping that will describe what’s going on for a minimum sample size and cheapest cost. Mervyn is a PhD student in Data Science & Analytics, whose research is supervised by Dr. Steven Prestwich and Dr. Armagan Tarim, in 2015. His main area of interest is Optimising Survey Design using various machine learning techniques.