Although there are numerous reasons for real estate analysts to construct spatial housing submarkets, there is little clarity about how this might best be done in practice. The existing literature offers a variety of techniques including those based on principal components analysis, cluster analysis and a range of other statistical procedures. This paper asks whether, given their market expertise and their role in disseminating information, shaping search patterns and informing bid formation, real estate agents might offer an effective but less data intensive method of submarket construction. The empirical research is based on an experiment that compares the predictive of different sets of submarket boundaries constructed by using either standard statistical methods or through consultation with real estate agents and other market analysts. The analysis draws on housing transactions data from Istanbul, Turkey. While the results do not demonstrate the outright superiority of any single method, they do suggest that expert-defined boundaries tend to perform at least as well as alternative construction techniques. Importantly, the results suggest that agent-based methods for delineating submarket boundaries might be used with a degree of confidence by real estate analysts and planners in market contexts where rich micro-datasets are not readily available. This has been one of the constraints internationally on wider adoption of submarket boundaries as an analytical tool.