Junior Data Scientist at a Manchester retailer: a public dashboard analytics directors search for
Four months. Five applications. One offer. The slowest job search I've ever run, and the one I'm most proud of. I was a PhD candidate in statistics in Manchester. I'd been working on my thesis for about two years when I decided I needed to find commercial work. Partly because the funding situation had changed, partly because I'd realised, fairly late, that I wanted to leave academia. I had no commercial experience. I had three published papers, one of which had been cited fifty times in a relatively small sub-field. I'd taught two undergraduate modules. I'd built a small simulation framework in R that nobody other than me used. The CV looked like a CV for a postdoc, not for a junior data scientist. Cold emailing hadn't worked when I'd tried it for a week. I'd sent twelve emails to data teams in Manchester and London and gotten one polite reply. The thing that worked was a public dashboard. I'd been wanting to learn dbt and Streamlit. I picked a topic, UK retail footfall, using publicly available data from one of the data trusts, and built a dashboard over six weekends. It wasn't elegant. It worked. I put it on a free Streamlit hosting tier and tweeted about it once. About two months later the head of analytics at a mid-sized retailer in Manchester found the dashboard via a Google search while researching a specific footfall topic. He didn't contact me through the dashboard. He searched for my name, found my LinkedIn, and sent me a message asking if I was open to a junior data scientist role. The interview process was three rounds over five weeks. Round one was a technical screen with two senior analysts. They gave me a piece of their data, a retail transaction dataset, and four questions. I had a week. I sent back a six-page Jupyter notebook with the answers, three caveats, and a writeup. They'd been hoping for code rather than narrative. I gave them both. Round two was a case study presentation. The brief was to take one of the answers from the take-home and turn it into a fifteen-minute presentation for a non-technical stakeholder. I rebuilt one section from scratch, simplified the visualisations, and presented to a panel of five: three analysts, the head of analytics, one product manager. The questions afterwards were specifically about what I'd cut and why. Round three was a team chat with three other junior data scientists who'd be my peers. Less of an interview, more of an "is this person tolerable to work with" check. The offer was £35,000 base, the higher end of the junior band. I asked for £38,000, citing the four months of search, the depth of the process, and a slightly higher quote I'd received from a competing offer. They came back at £38,000. I accepted. I'd sent five formal applications in total. The one that worked was the one I hadn't really applied to.