INTRODUCTION: Health information management system data is collected for national planning and evaluation but is rarely used for healthcare improvements at subnational or facility-level in low-and-middle-income countries. Research suggests that perceived data quality and lack of feedback are contributing factors. We aimed to understand maternity care providers' perceptions of data and how they use it, with a view to co-design interventions to improve data quality and use. METHODS: We based our research on constructivist grounded theory. We conducted 14 in-depth interviews, two focus group discussions with maternity care providers and 48 hours of observations in maternity wards to understand maternity providers' interaction with data in two rural hospitals in Southern Tanzania. Constant comparative data analysis was applied to develop initial and focused codes, subcategories and categories were continuously validated through peer and member checks. RESULTS: Maternity care providers found routine health information data of little use to reconcile demands from managers, the community and their challenging working environment within their daily work. They thus added informal narrative documentation sources. They created alternative narratives through data of a maternity care where mothers and babies were safeguarded. The resulting documentation system, however, led to duplication and increased systemic complexity. CONCLUSIONS: Current health information systems may not meet all data demands of maternity care providers, or other healthcare workers. Policy makers and health information system specialists need to acknowledge different ways of data use beyond health service planning, with an emphasis on healthcare providers' data needs for clinical documentation.