I've been running or participating in calibration sessions for about 8 years. most engineers have no idea what happens once their packet goes in. here's the real version.
the format
it varies by company but the common version: a set of managers and their skip-level (often a director or VP) get in a room or video call. each manager presents their promo candidates. the group discusses and either ratifies or challenges. there's usually a budget: X promotions approved at this level, across this org.
what your manager actually says
if they've prepared well, they open with a 30-second case statement and then read 2-3 specific impact highlights from the packet. the goal is to get a 'yes, that tracks' from the room and move on.
if they haven't prepared well, they say 'she's really been stepping up' and then stumble when someone asks for specifics. that's how you lose in calibration even with a strong packet.
the budget constraint is real
this is the part nobody tells ICs. if your org has 40 senior engineers and the promotion budget allows 3 staff promotions per cycle, you can do everything right and still not make it. not because you weren't ready. because the math didn't work.
the workaround some people discover: your manager can sometimes push harder if you're 'cleanly' above the bar, but they can't manufacture budget that doesn't exist.
cross-functional credibility
peer feedback from PMs, data scientists, or design partners often carries disproportionate weight at senior-to-staff transitions. engineers who only get endorsements from other engineers on their immediate team are seen as operating in a smaller sphere. the calibration room notices.
what you can actually control
ask your manager before cycle: 'how many names from our org are going in?' and 'where am I in that stack?' they may not tell you everything but the question signals you understand how this works and they usually give you something.