This article discusses improvements to the methodology of the Housing and Urban Development (HUD) point-in-time (P-I-T) homeless census with an aim to capture correct numbers of people experiencing homelessness. HUD’s P-I-T results are presented to Congress as official data for policy consideration. Yet, PIT methodology focuses on visible street homeless individuals and those in shelters while neglecting the “marginally housed” or less visible homeless who live in automobiles or temporarily stay with friends and extended family. Being a hidden population, the marginally housed has been a traditionally difficult population to study, HUD’s PIT count was replicated and additionally targeted the marginally housed to improve traditional methods of counting the homeless.

The P-I-T count was improved in two ways:

(1) By extensively training counters

(2) By using the personal networks of hundreds of counters to seek out the marginally housed.