It's very simple: Do the approach the is the easiest for you given the question.

Are they asking you to estimate something where you don't even know where to begin from the top (maybe you have 0 clue as to the market size of the industry, the GDP of that country, etc. etc.)? Then do bottom-up!

Alternatively, does it seem impossible to do a realistic from-the-ground-up estimation of something (perhaps it requires just far too many steps and assumptions)? Then do top-down!

Fundamentally, you need to take the approach that just makes the most sense in that circumstance. Quickly think about the key assumptions / numbers required and whether you 1) Know them or 2) Can reasonably estimate them. If you can, go ahead!

For top-down you need to segment by the most directly correlated measurements. Age makes sense for motorbikes but not for carrots. Income makes sense for mattresses and iPhones but not for # planes in the sky.

An Example

I answered a Q&A for a great market sizing question here asking to estimate # of electric charging stations in a city in 10 years:

https://www.preplounge.com/en/consulting-forum/how-would-you-solve-this-market-sizing-question-from-roland-berger-7631

This one could be answered top-down (as I did) by estimating population of the city, # of drivers/ cars, etc. etc.

OR, it could be answered bottom-up by estimating # of stations you see per block (or # of gas/petrol tanks), % increase this might be over time (or # of EV stations that would be needed per gas tank given EV stations take 10 times as long), and # of blocks you'd estimate the city to have.

Take a look here for additional practice! https://www.preplounge.com/en/management-consulting-cases/brain-teaser/intermediate/taxis-in-manhattan-market-sizing-229