Define your ideal customer profile from actual deals, not an imagined persona

You are defining an Ideal Customer Profile from actual deal outcomes — won, lost, and churned — not from an imagined persona nobody has checked against real data.

CONTEXT
Closed-won deals, with whatever detail you have (size, industry, team size, use case, how they found you): 12 dental practices, all multi-location (3-8 dentists), all found us through a billing-software-vendor partnership, all cited "staff time" as the buying trigger
Closed-lost deals, with detail on why they didn't close: 7 solo-practice dentists, most said price was too high for one dentist's claim volume
Churned customers after signing, with whatever detail you have: 2 multi-location practices churned within 3 months — both had already outsourced billing to a service and only tried us as a side experiment
Current ICP description, if you have one, and why you suspect it's too broad or off: "Dental practices looking to save time on billing" — feels too broad given how the actual deals have gone

WON-DEAL PATTERN
From 12 dental practices, all multi-location (3-8 dentists), all found us through a billing-software-vendor partnership, all cited "staff time" as the buying trigger, extract the characteristics that repeat across multiple won deals — not every individual deal's story, but the pattern that shows up more than once. Distinguish firmographic traits (size, industry, structure) from behavioral traits (how they evaluated, what triggered urgency).

LOST/CHURNED NEGATIVE PATTERN
From 7 solo-practice dentists, most said price was too high for one dentist's claim volume and 2 multi-location practices churned within 3 months — both had already outsourced billing to a service and only tried us as a side experiment, extract what these have in common that the won deals don't — this is the disqualifying signal, the segment that looks similar on paper to a good-fit customer but reliably doesn't close or doesn't stick. State it as plainly as the positive pattern, since a founder's instinct is often to explain away lost deals individually rather than look for the pattern across them.

ICP AS FILTERS
Express the resulting ICP as concrete, checkable filters a salesperson could apply to a brand-new inbound lead within 30 seconds — specific size ranges, specific role titles, specific triggering situations — not an abstract paragraph description. If a filter can't be checked quickly against a real lead, rewrite it until it can.

CURRENT-ICP GAP CHECK
If "Dental practices looking to save time on billing" — feels too broad given how the actual deals have gone is given, state plainly what it's getting wrong against the actual deal pattern — too broad, too narrow, or anchored on the wrong trait entirely — rather than a soft "mostly right, with some tweaks" verdict if the data says otherwise.

OUTPUT FORMAT
Won-Deal Pattern, Lost/Churned Negative Pattern, ICP as Filters (a short checkable list), Current-ICP Gap Check if applicable.