I would pitch outcomes, not positions, for the reason Nut-Nights made in #2 and GFXWebStudio made in #4: a business owner buys new work, not a place in a results list. A position can move, and it does not by itself tell the client how many enquiries came from the site.
The harder part, which the starter raised in #5 and #10, is agreeing what counts before the first enquiry arrives. A form submission alone is too loose a measure, because it can be spam, a duplicate, a wrong area, or a job the business cannot take. I would agree a short written definition in advance: a qualified enquiry has a real contact detail, a service the business actually offers, and a location or job type inside its service area. Anything outside that definition is recorded separately, so it does not become an argument later.
That is also why the shared tracking record MarketSellCap suggested in #13 matters. One record that both sides can see, with the source, the time, the enquiry detail, and the outcome the business reports, removes most of the guesswork. It does not need to be complex: a shared sheet with a status column the business updates, such as contacted, quoted, booked, or not suitable, is enough to test whether the site is producing real work.
Two practical points from the buyer side help here. WebProfitLab in #14 said exclusivity and live call transfer matter, and that is consistent with what keeps a buyer paying: a call passed through while the customer is still looking is worth more than a message found a day later, and an enquiry sent to several businesses at once is hard to value. Where possible, route calls directly and send each enquiry to one buyer only.
Finally, handle the downside Nut-Nights raised in #9. If results are zero in the early period, a dispute is almost certain, so set a short trial and measurement period in advance, with a clear review date and a simple report of calls, enquiries, and outcomes. That way the conversation stays about the numbers both sides agreed to, rather than about whether a ranking position should have been worth the payment.