Your best intern left for the competitor in week six
Booked to shadow for ten weeks, she ran the biggest launch by week four and had a competitor's offer by week six. What an optics-built internship really costs.
By week 4, the intern was running the product team's biggest launch of the quarter. She was a third-year computer-science student from a state school the company had never recruited. Engineering leads wanted her read on the roadmap. By week 8 she had shipped more measurable impact than two of the junior engineers beside her.
By week 10 she had two competing offers. The company had made her none. Its program existed to build a pipeline, which in practice meant staying friendly with campus recruiting and hoping the strong ones came back in a year. She did not come back. She took the offer a competitor extended in week 6. The company had paid for the training. The competitor kept the hire.
This is the common failure mode of a growth-stage internship. The program runs for marketing. It generates recruiting posts, campus credibility, the feeling of investing in the next generation. It does not run as a hiring pipeline, because nobody built the machinery to convert one. The strong intern gets used as a cheap engineer for a summer, then handed back to the market with sharper skills and no offer.
The math is unkind. At one company I worked with, a single summer intern cost roughly thirty thousand dollars once you counted salary, equipment, and senior-engineer mentoring hours. That spend buys one thing worth having: the option to hire someone you have already watched perform. Let her walk and you keep the cost and lose the option. A competitor then lands a candidate with a proven record on shipped work, at the price of an ordinary new-grad offer.
Companies that convert treat their top interns as accelerated hires. By week 4 a manager has flagged who is exceptional. By week 6 the conversation has happened, informally but plainly. We want you back. Here is roughly what the offer looks like. Here is what you would own. Wait until the final week and your offer lands third or fourth on her desk, in a negotiation she now controls. Move early and you take yourself out of that auction. The same discipline that makes a comp band worth defending applies here: decide the number before the market sets it for you.
Design the program backward from conversion. Named managers accountable for each intern. Real ownership that produces evidence of capability. A scheduled conversion talk, not a hopeful one. The care a good team puts into onboarding a new hire is the care a serious program puts into keeping its best intern. The one who ran the real project was the one you most needed to keep. Treat her as a hire from week one, or train her for someone else.