Learning to spell in English requires the integration of general and specific word knowledge. This paper describes the 'Word Nerds' project, a research–practice partnership consisting of two researchers from a large public university and 17 elementary teachers in seven school districts in the United States. The collaboration was formed to study variation in instructional practice among teachers using the Words Their Way programme and address teacher‐generated questions related to how children learn to spell words. This paper describes how stakeholders worked together to understand (1) what grouping and organisational structures teachers use for spelling instruction, (2) the extent to which elementary students use analogy to spell unknown words that share spelling patterns with known words and (3) the affordances and challenges of a partnership approach to educational research. Data collection took place over one school year. Data sources included teacher focus groups, classroom observations and student formative spelling assessment data (n = 178). Analyses included descriptive statistics, t‐tests and multilevel modelling, nesting spelling items in students in classrooms. Findings indicated that classroom observation scores for grouping, teacher talk, student‐to‐student talk, reflection and student engagement during spelling lessons varied depending on format (whole group, small group or pair/individual). Students showed evidence of spelling by analogy, but this strategy was associated in part with the semantic difficulty of the target words. Instructional practices for spelling vary based on teachers' priorities, and teachers need support to implement and prioritise evidence‐based practices. Conducting research in partnership allowed the researchers to observe and understand variation in implementation of a popular spelling programme and to develop and answer research questions that mattered to teachers; however, it also limited the size, scope and location of the project. Highlights: What is already known about this topic Children can read and spell words by analogising familiar spelling patterns from known to unknown words.Evidence‐based spelling instruction focuses on common orthographic and morphological patterns that help students use analogy to spell and form high‐quality lexical representations in memory.Implementation of evidence‐based spelling instruction is challenging for teachers, and there is little research examining how popular commercial spelling programmes are used in classrooms. What this paper adds Classrooms engaged in small‐group spelling instruction included more teacher talk; classrooms engaged in individual instruction were highly differentiated; and classrooms engaged in whole‐group instruction included more student‐to‐student talk and engagement.Student data showed evidence of spelling by analogy, but students were less likely to spell more semantically challenging words correctly and made more errors in spelling those words, even when orthographic difficulty was held constant.The research–practice partnership allowed researchers to document and explain variability in spelling instruction and address teachers' questions about spelling development. Implications for theory, policy or practice Teachers adapt spelling instruction to address their priorities, and different organisational models prioritise direct instruction, differentiation, standards‐based content, or peer collaboration.The ability to analogise spelling patterns from known to unknown words may depend on linking orthographic knowledge to vocabulary knowledge.Teachers need support and community to implement and iterate best practices for spelling instruction. [ABSTRACT FROM AUTHOR]