This article originally appeared in American Banker.

For years, banks have relied on lean techniques to improve branch networks and processes like mortgage and credit-card applications. Their overarching goals have been to reduce cost and raise productivity. Yet two-thirds of lean programs either don't deliver the desired level of cost reduction or can't sustain the savings, according to a Bain & Co. assessment of 17 financial institutions.

One reason, we would argue, is banks' heavy reliance on testing a new process through small pilot projects. Banks would gain a much better understanding of how new programs and initiatives will work on a broad scale by embracing hothouse testing. A hothouse test environment focuses on average processes, average customers and average locations—and it uses a sufficiently large sample with enough variation among these variables to represent the entire network.

In practice, it's difficult to draw useful lessons from a small pilot. Banks typically stack the deck for success. They pick one or two branch or back-office locations with the best and most motivated managers, the easiest problems to address and the most receptive customers, and they ensure adequate staffing for the project.

As a consequence, small pilots often fail to subsequently roll out at large scale. They can rarely withstand a more complex process, average managers or less receptive customers. When the main unit of cost consists of labor, improvements to a process at one branch have a tiny effect and become difficult to expand to a meaningful size.

Read the full article at American Banker.

Peter Stumbles and Richard Fleming are partners with the financial services practice of Bain & Co.