At first, it seems like a great idea. But then you really think about the incentives in place, and they aren't really working in your benefit.
If the candidate doesn't have to pay, that means that every candidate is happy to have a session. There's no risk for them. They may not be willing to put in the effort, they may know they have no chances at all, but what do they lose by doing a free coaching session? Nothing. Some candidates need 1-2 seesions, other need 3-4, and there's the occasional candidate that wants to use the coach as the only person who they do cases with…
So, if you are a coach and you want to maximize your profit, you need to “veto” first your candidates. Candidates that are sure to get the job. Candidates that don't need much work. The better you “veto”, the better you do. You'll actually do better by spending extra time veto'ing candidates than by just coaching them.
This is exactly how the model you suggest has been implemented. You apply, they check if you are high potential, and if you are, they'll take x% of your annual salary. They select candidates that don't need much coaching, and are highly commited, and then take a % of their salary. Then they advertise how successful they are in getting offers… but their success is not in the coaching process but in the “selection process”. I suggest running away from that model as fast as you can. Your incentives are simply not aligned.
Nevertheless, risk sharing is a good idea (but what you suggested is not risk sharing, but instead risk transfer from the candidate to the coach). For example, having a base price for a session + extra in case of success could be a reasonable model. However, it is not how Preplounge works, and has the logistics problems (credit risk, time of payment…) that others mentioned.