I have a service which identifies which accounts will pay for accounts receivable. IE: collections agencies. Numbers show it can move the needle double to triple digits depending on the account mix.
Right now, I'm planning on a per seat monthly subscription. Customers know how much they'll be charged each month, and I think it lets me focus on the value rather than community unit costs.
EDIT: Just a note, unfortunately, this is a very high touch environment. The upshot being that once a customer chooses you, they never leave. Even if they're being outright abused. (I have stories!)
I believe it was Patrick who recommended that on Twitter the other week.
If you're really getting X0% to Y00% in per-collector productivity then I say start by asking for for $100 per seat per month, if you're wedded to per-seat pricing. You can price anchor it around "less than a single collector-day" and that leaves plenty on the table for your clients. I'd then start walking the price up (if it is worth $100 a collector it is probably worth $150/$200/$250) until you start losing lots of sales on the pricing question.
If you've got very heterogenous clients number of people on the phones might not capture value as well. You might consider doing N tiers (maybe or maybe not publicly disclosed). I'd be thinking something along the lines of $500 for the mom&pop shops, $2.5k for "real businesses" doing debt collection, and $10k++ custom pricing for large/enterprise accounts. You could have multiple segmentation levers between plans, such as e.g. size of the average receivable, total volume of receivables, type of collection, what data sources you pull in to do the calculations ("Pay more get more!"), etc.
The great thing about high-touch sales is that if you hate how your pricing strategy interacts with any one client you just change it for the next one.
If you'd like to discuss this in more detail non-publicly, email me. I have a weird personal interest in this field and am happy to trade informal consulting on SaaS pricing for war stories.