Experimental Design
We collected an in-person baseline survey in Jan--March of 2023, with a total sample of roughly 1000. We then did an 8-month follow-up to construct a panel dataset of key shop-level outcomes. Using this panel we construct a simple measure of gross profit/value added on a set of fixed effects to control for time and location. Within each "industry" (defined by the main products sold by the shop) we identify the most profitable 15% as "mentors" and assign the rest to pool of potential mentees. Within each location and industry, we randomly select roughly 300 for the intervention, and randomly select 300 of those untreated as our control group. We then assign mentees to mentors within their industry but outside of their location using sequential randomization. Our partner Prospero then does the intervention: mentees are taken to the mentor's shop for a facilitated in-person session where they are guided to share advice and observe the mentor's practices. After the intervention we will collect a second round of in-person data, and then another 3 months of follow-up.