this is one of the things you learn in an MBA program. you'd generally use multiple methods to triangulate a rational estimate. for example, you'd look at industry comparables and estimate a beta to plug into CAPM (as sibling commenter touches on). you can also do full financial projections (5-7 years out) based on expected performance, do DCF, and monte carlo that to back out a discount rate. you can also do a comparative ratio analysis (https://www.investopedia.com/terms/r/ratioanalysis.asp) on profitability, cash flow, asset efficiency, turnover, and the like.
from my (limited) experience, startup valuations seem to be most sensitive to growth rate and the discount rate, so modelers spend a lot of time estimating these factors.