Over the last decade there have been numerous studies on touchscreen typing by blind people. However, there are no reports about blind users- everyday typing performance and how it relates to laboratory settings. We conducted a longitudinal study involving five participants to investigate how blind users truly type on their smartphones. For 12 weeks, we collected field data, coupled with eight weekly laboratory sessions. This article provides a thorough analysis of everyday typing data and its relationship with controlled laboratory assessments. We improve state-of-the-art techniques to obtain intent from field data, and provide insights on real-world performance. Our findings show that users improve over time, even though it is at a slow rate. Substitutions are the most common type of error and have a significant impact on entry rates in both field and laboratory settings. Results show that participants are 1.3-2 times faster when typing during everyday tasks. On the other hand, they are less accurate. We finished by deriving some implications that should inform the design of a future virtual keyboard for nonvisual input. Moreover, findings should be of interest to keyboard designers and researchers looking to conduct field studies to understand everyday input performance.