When looking at the growth rates of the business intelligence platform space, it is apparent that acquisitions of new business intelligence tools have shifted dramatically from traditional data visualization and aggregation use cases to newer data discovery implementations. This shift toward data discovery use cases has been driven by two key factors: faster implementation times and the ability to visualize and manipulate data as quickly as an analyst can click a mouse. The improvements in implementation speeds stem from the use of architectures that access source data directly without having to first aggregate all the data in a central location such as an enterprise data warehouse or departmental data mart. The promise of fast manipulation of data has largely been accomplished by employing in-memory data management models to exploit the speed advantage of accessing data from server memory over traditional disk-based approaches. The “physics” of data access favors in-memory data management models. However, in-memory techniques are not without drawbacks. As companies attempt to evolve from small departmental projects to broader division-wide or enterprise-wide initiatives, increasing data volumes and the impact of increasing data consumer counts challenge the limits of early in-memory implementations. These challenges raise serious questions that should be considered by any organization considering in-memory techniques for business intelligence platforms. Avoiding the Barriers of In-Memory Business Intelligence: Making Data Discovery Scalable