Hiring Education Data Talent: What We've Learned (Part 1)
This is part of a series sharing what we've learned in more than 15 years of attracting and screening education data talent, and what it means in meeting the moment of today's hiring market.
A former partner recently called me in full panic. They had posted a data analyst role expecting a few dozen applicants. Instead, they got 392.
On paper, it looked like a gift. But when they gave semi-finalist candidates a basic analytic task, the results were sobering. Many of the applicants who looked strongest on paper turned out not to have the depth of the skills their applications suggested. The résumés were perfect. The real abilities were less so.
We see versions of this story all the time, and we have learned too much from it to keep it to ourselves.
AI can now generate a flawless résumé and a sharp cover letter. What it cannot show is who can sit in a room with a superintendent, a budget director, or an educator and move a real decision forward. It cannot show who can take a messy dataset, find the signal, and explain it in plain English to people who do not think in regressions. It cannot show who can walk in with skills in one programming language and quickly learn another, or help garner enthusiasm for a new, better way to address a problem. It cannot show who will keep going when a project meets the perils of politics, bad data, or radio silence from leadership.
Those are not “soft” skills. They are the core of the job.
In public education, the stakes are especially high. A weak hire is not just an inconvenience. It can mean missed reporting deadlines, policy decisions made on the wrong numbers, or leadership losing confidence in data altogether. A bad hire in a data office can mean FAFSA errors, misreported graduation rates, or districts making staffing decisions off the wrong numbers.
After hiring, training, and watching hundreds of data fellows grow inside real agencies, we’ve learned a few things about hiring that we think are useful for the whole field.