To be even thinking of ROI when speaking with someone will almost certainly kill off any chances of something coming from it. Your attention needs to be focused on them rather than what benefit you may or may not get from the discussion.
To assume that you cannot measure the ROI of social interactions, however, is not totally true. Some things are hard to measure but that does not mean they cannot be measured.
Measuring one tiny feather may be difficult. Measuring a large collection of feathers, however, is much easier. It’s like the old joke ‘How much does a ton of feathers weigh?!’ A single conversation, when selling services, will rarely lead to a sale. Several conversations over time – maybe days, weeks, or even years – could be totally different. Which conversation made the difference? The first? the last? who can tell? Grouping the interactions together over time makes measurement much easier.
Social interactions on platforms like Twitter are harder but still possible. The key is to track results over time. Change your approach and start tracking. You probably will want to give it a reasonable period of time, such as a couple of months, but you should be able to compare cost of time invested with results back. You may measure in terms of cash, hits to website, or some other way. What matters is that over time you can see results that originated from your Twitter activity. You can then compare them with the previous results over a similar period.
A top door-to-door sales person I once modelled would change her pitch slightly every day and compare the results to her previous average. If it were better she would keep the new pitch. If less she would go back to the previous one. One use of the new pitch would not mean anything. After knocking on 100 doors she started to get the picture.
I am currently doing work with a big marketing team, helping them to measure the ROI of various forms of marketing activity. The process is actually very simple. You set up a campaign, do the activity, and then measure the leads generated and resulting sales compared to the costs. As long as you have the mechanisms to capture the information the maths is easy! If you were doing an email marketing campaign you could even measure the average ROI of each email sent.
The point is that if working out the ROI is important to you, it can be done but you have to do it in hindsight and you have to start tracking results and comparing them with the activity and costs invested.