Purpose: The purpose of this project was (1) to collaborate with public preschool programs to adapt, test, and revise the implementation model of an innovative early mathematics intervention, "Pre-K Mathematics" and (2) to determine whether this model makes it possible for local programs to sustain their implementation without reducing the intervention's effectiveness. This intervention has been evaluated in multiple RCTs and has been found to be effective by the What Works Clearinghouse. In the RCTs, the intervention was implemented in a standard way -- using a standard implementation model -- for a school year. This model ensures that teachers provide essential active ingredients, such as curriculum dosage and aspects of instruction, throughout the school year. In following up with teachers in subsequent school years, we have found that some teachers' fidelity to the standard implementation model drifts, as teachers scale in (i.e., adapt) the implementation model. A consequence of scaling in can be fidelity drift, by which active ingredients of the intervention are lost, which diminishes the effectiveness of the intervention. In this project, a continuous improvement process was used to scale-in "Pre-K Mathematics" in a sustainable way without reducing its effectiveness. Setting: The project was conducted in 5 LEAs in California, which varied structurally (school district and county department of education), geographically (urban and rural), and in racial/ethnic composition of communities served. Improvement Approach: The framework for our theory of action combines features of two distinct management approaches: innovation and Kaizen (continuous improvement). The "Pre-K Mathematics" intervention, with its reliance on intentional, small-group mathematics activities for the pre-K classroom and dyadic activities for the home, tracking of dosage delivery, authentic assessment and monitoring of all children's performance and progress, is an innovative approach to supporting early math in pre-K classrooms. The partnership collaboratively adapted the standard implementation model, which resulted in a modified implementation model. This model was tested, and further modified when warranted, in short cycles for one year. A Sustainability Study was then conducted to determine whether LEAs using the modified model could sustain an effective implementation of "Pre-K Mathematics" for a full school year. Project Participants and Design: The 65 lead teachers (41 in Cohort 1 and 24 in Cohort 2), who participated in this CI study, were previously trained to use the standard implementation model (SIM) for "Pre-K Mathematics" during their participation in the RCT. These teachers/classrooms were re-randomized into two conditions. Blocking on local program, half were randomly assigned to the Continuous Improvement condition (use of the MIM) and half to the comparison condition (SIM fidelity drift), resulting in 34 CI teachers/classrooms and 31 comparison teachers/classrooms. The 669 pre-kindergarten children (431 in cohort 1 and 238 in cohort 2) comprising the research sample for the comparison study were randomly selected within each participating classroom. Twenty-one lead teachers/classrooms and 176 pre-kindergarten children from the original RCT comprised the Business-As-Usual control condition. Continuous Improvement, Comparison, and BAU Control Teachers: Continuous improvement and comparison teachers had been trained to use the standard implementation model for "Pre-K Mathematics" during their participation in a prior RCT. In the present project, continuous improvement teachers collaboratively developed and implemented a modified implementation model. Comparison teachers continued to implement "Pre-K Mathematics" without engaging in continuous improvement, which allowed their fidelity to the standard implementation model to drift (SIM fidelity drift). Control teachers did not implement "Pre-K Mathematics." Key Measures: Implementation data used in short cycle testing and the sustainability study included (1) intervention fidelity, (2) classroom curriculum dosage by teachers, (3) home curriculum dosage by parents, and (4) child math mastery (progress monitoring). In the sustainability study, children's mathematical knowledge was assessed at the beginning (pretest) and end of the pre-k year (posttest) using the CMA. Findings and Conclusions: Short Cycle Testing. CI teachers' levels of fidelity, curriculum dosage delivered by the teacher, curriculum dosage delivered by parents, and children's math mastery scores were compared to criterion levels using chi square. The partnership tested four modifications of the implementation model. Three were retained: (1) train instructional assistants to help the lead teacher with implementation, (2) include more time in the curriculum plan for children who need extra help or who have been absent, (3) implement a family engagement strategy to strengthen support for math at home (e.g., a family math event at the beginning of the school year). The fourth modification was not retained: (4) increase the size of small groups for classroom math instruction. Sustainability Study. The quality of implementation was found to be significantly higher in CI teachers than in fidelity drift teachers (Table 1). CI teachers delivered the "Pre-K Mathematics" curriculum with significantly higher fidelity. They provided more curriculum dosage in the classroom. Parents of the children in CI classrooms provided more curriculum dosage at home. Consequently, CI children mastered more math activities (i.e., engaged in independent problem solving without help). Children's math outcomes differed significantly by condition (Figure 1). A two-level ANCOVA was conducted with children nested within classrooms, and repeated observations (pretest, posttest observations) on children, and with the pretest score and pretest age as covariates. The pretest math score by condition interaction was significant. When pretest math scores were low (one SD below the mean), CI treatment and fidelity drift children, both of whom received the "Pre-K Mathematics" curriculum, outperformed control children. When pretest math scores were average, all groups differed, with CI treatment children performing best and control children worst. When pretest math scores were high (one SD above the mean), CI treatment children outperformed fidelity drift and control children. Math outcomes were higher for CI treatment children than for control children, ES=0.69. Math outcomes were also higher for drift children than for control children, ES=0.41, indicating that "Pre-K Mathematics" remained effective to a diminished degree under drift conditions. In summary, after teachers modified their implementation in a manner that made it more sustainable, "Pre-K Mathematics" continued to have a beneficial impact on children's mathematical knowledge. Our empirical approach to continuous improvement enabled local programs to sustain their implementation of the intervention without reducing effectiveness.