The purpose of this study is to present the need for stakeholders to share an agreed-on Theory of Change (ToC) during the various phases of an implementation process. To this end, we draw on the Swedish Boost for Mathematics (BM), a large-scale national educational initiative, and its intrinsic 'algebra' instructional module. More precisely, we analyze a cohort of official evaluation reports on the BM from the perspective of a critical case study. The BM's embedded algebra module and its associated material offer an instrumental example. This analysis relies on a selection of theoretical constructs related to Implementation Research (IR), such as (educational) innovation, value, stakeholders, degree of merit, policy dimension, theory-driven evaluation, and ToC. These theoretical constructs derive from both general IR and mathematics education research. We identify four phases of implementation, based on the analyses of the BM, namely, policy and organization, development, implementation, and evaluation. The analyses illustrate the need for alignment among these phases, associated with the endorsed, enacted, recognized and applied ToC of stakeholders. As a critical example, BM illustrates the many pitfalls of large-scale, scaled-up educational innovations--one of these being the potential failure of evaluations--owing to lack of an agreed-on ToC.