Payment for performance (P4P) initiatives have been employed in low and middle-income (LMIC) countries as a means to improve the delivery and coverage of maternal and child health (MCH) services. Despite widespread implementation, there is still a lack of consensus on whether P4P is an effective initiative that leads to positive, sustained improvement in delivery of these services. There is a need to employ methods that can evaluate the pathways through which P4P alters health systems without diminishing the complex behaviour exhibited by health systems in the evaluation. Two methods for evaluation of complex systems were used to model the impact of a P4P programme on the delivery and uptake of MCH services in Tanzania: causal loop diagrams (CLDs) and system dynamics modelling (SDM). The CLD represents relationships between variables that are important when we consider how the health system responds and transforms under P4P. The CLD was developed using qualitative data from a process evaluation of a P4P programme in Tanzania and stakeholder consultation. The CLD was then used to build a quantitative SDM, using primary (stakeholder consultation) and secondary (impact evaluation of P4P programme, official statistics and reports) data sources. In the SDM, changes in design, implementation, and context (availability and supply of drugs, access to alternative sources of funding, staffing) were tested to explore the impact on key outcomes (percentage of women who received two doses of intermittent preventive treatment during antenatal care and percentage of women who had a facility-based delivery) and the effectiveness of the programme. The CLD pinpoints the key mechanisms underpinning provider achievement of P4P targets, reporting of health information by providers, and care seeking by the population, and identifies those mechanisms affected by P4P. For example, the availability of drugs and medical commodities was critical not only to provider achievement of P4P targets (supply of MCH services) but also to demand of services and was impacted by P4P through the availability of additional facility resources. In the SDM, severe delays in payment and change in allocation of payments (between staff and operations) impacted key outcomes, with changes in contextual factors (particularly provision of medicine) facilitating or hindering facility performance. Recommendations for programme design must consider the impact on the holistic system, to avoid suboptimal programme impact or unintended, negative consequences. Our study shows how secondary data from an impact and process evaluation can be used to model the health system and its response to P4P, to improve our understanding of programme mechanisms and inform the design of more effective future P4P programmes. This work will not only be relevant for P4P in Tanzania but also generate policy relevant recommendations for LMICs.