Background/context: Student heterogeneity is a persistent and fundamental challenge in all school systems. In many countries, including Norway, politicians and actors within the educational sector have increased the use of additional teachers, teacher assistants, and special education teachers to respond to this challenge (Bonesrønning, Iversen & Pettersen, 2011; Kirkebøen, Kotsadam, Raaum; Andresen & Rogstad 2017; Reiling, Salvanes, Sandsør & Strøm, 2021, Sandsør, Reiling & Salvanes, 2019). However, there is still little knowledge on how to use teachers and other professional staff efficiently to increase student learning (Solheim & Opheim, 2019). Typically, research has focused on the effect of reduced class size or increased teacher or assistant density where findings tend to show small or no effects on student learning (see e.g., Blatchford, Russell & Webster., 2012; Leuven, Oosterbeek & Rønning, 2008; Iversen & Bonesrønning, 2013; Falch, Sandsør & Strøm, 2017; Leuven & Løkken, 2018). These results have motivated a search for alternative ways of organizing teaching, where tutoring and tracking are among the proposed means (Blatchford et al., 2012; Dobbie & Fryer, 2013; Betts, 2011). Individualized, intensive, in-school tutoring has been shown to improve student performance (Fryer 2017; Guryan et al. 2021), as have early interventions (Carneiro & Heckman 2003). This suggests that using additional teachers to provide small group instruction in early grades may be a particularly efficient way of organizing teachers in order to increase student performance. Purpose/Objective/Research Question: Using a large randomized eld experiment, unprecedented in scope and duration, we investigate the effect of small group instruction on mathematics scores. The main objective of the study is to investigate whether students gain from individualized small group instruction and whether there is heterogeneity in treatment effects (The 1+1 Project, https://1pluss1prosjektet.no/frontpage). Setting: The eld experiment is carried out in Norwegian public primary schools, grades 2-4 (ages 7-10). Compulsory education is comprehensive, with a common curriculum and no tracking and is nearly entirely public and provided free of charge. Grade promotion and retention is uncommon, ensuring that nearly all students follow their cohort Students take mandatory tests in reading and mathematics in 5th grade, where results are conveyed to teachers and parents but have no direct consequences for students. Population/Participants/Subjects: The project includes 160 Norwegian primary schools, located in 10 large municipalities geographically spread across the country, equally divided between a treatment and control group in a strati ed randomization. Four adjacent cohorts of students have participated in grades 2-4 for 1, 2, or 3 years (see Table 1). Intervention/Program/Practice: Treatment schools receive an additional teachers, used to provide small group instruction in mathematics. Students are pulled out of their regular classes for two periods of 4-6 weeks during the school year and offered mathematics instruction 3-5 times a week in small groups of 4-6 students. Small group instruction is carried out in parallel to regular instruction, covering the same topics but with the potential of providing more tailored instruction in smaller groups. All students are targeted by the intervention. The majority of small groups consist of students at similar academic achievement levels. Research Design: Using a stratifed randomized eld experiment design, we investigate the effect of small group instruction on 5 grade national test scores in mathematics, controlling for background characteristics. National test scores are from administrative data, eliminating or greatly reducing bias from consent and attrition. For a subset of consenting students (80-90%) we are able to combine register data with project data, including yearly test scores and detailed registration of participation in small group instruction, allowing us to investigate mechanisms in greater detail. Data Collection and Analysis: The data from the intervention are unusually rich. The targeted students in treatment and control schools are tested in mathematics before and after the intervention. Information is collected uninterruptedly about small group composition and length of small group exposure. Register data provides student background characteristics and results on standardized tests in 5 grade for the entire population of Norwegian students. Teacher characteristics, teacher behavior and information about student behavior in class are provided from repeated teacher surveys and school organization and leadership information from school leaders' surveys. Case studies add to the rich survey data. Findings/Results: The preliminary results from the first two cohorts that completed the national test in 5th grade in the fall of 2018 and 2019 (see Table 2) show that students in treatment schools scored on average about 0.06-0.07 standard deviations better on national tests than students in control schools. The difference is statistically significant and in line with existing evidence from low dosage tutoring. Further, our analysis of end of year tests administered by the project immediately after receiving treatment reveal treatment estimates about twice the size. These first results suggest a positive average treatment effect on test scores. Moreover, there are indications that the treatment effects are heterogeneous, i.e., they differ across student subgroups and schools. Conclusion: Using additional teachers to provide small group instruction in mathematics increases student test scores. Importantly, these results are found in a context where general increases in teacher density have not shown effects on student learning, suggesting that how additional teachers are organized is vital for achieving such performance gains.