I keep repeating it: Market sizing is not about coming up with a correct number, it's about coming up with a plausible way to number.
So your initial instinct to consider various segments (for example households, workplaces, gastronomy) does make sense. Then you could assume that an average person consumes x cups per week at home, y at work and z in public places.
Or you could start with average coffee consumption per consumer and then break it down by channel.
Also you should specify what is meant by "coffee market" - the market for coffee ground and beans? Or including gastronomy?