Carepatron growth · 26 June 2026
Contents
Trigger: huddle with Jamie (2026-06-26). An ex-Lyft product-growth lead (referred by Vista, pitching for investment) signed up, used the product, and gave one piece of unprompted feedback: Carepatron is badly under-investing in referrals. His thesis: lovable, shareable, low-barrier product where every user knows other potential users; referral should be a major channel, not a footnote. At Lyft it was the primary growth engine (~70% of paid customers). His headline prescription: a B2B referral reward worth $1,000 to $2,000, ungate it to everyone, and make it "a product within a product."
This brief pressure-tests that against our own data and economics so the number is a decision, not a guess.
The single biggest unknown is volume elasticity -- how much referral volume a 3.3x reward unlocks. We've never raised it, so we can't prove it from our data. July is the test. Frame it that way.
I pulled the live PostHog funnel. It confirms the advisor's read with our numbers (last 6 months, test-filtered):
| Stage | Volume / month | Read |
|---|---|---|
| Open the referral surface | ~250 to 440 accounts | Real top-of-funnel interest exists |
| Actually send a referral | ~80 to 155 accounts | Big drop -- most who look don't act |
| Referred signups | ~10 to 20 | Tiny output |
| Referred paid conversions | not instrumented before May 2026 | We can't even see the bottom of the funnel yet |
For scale: we add ~300 to 410 new paid customers/month. Referrals contribute low-single-digit %. That's the signature of an under-incentivized channel -- interest at the top, almost nothing out the bottom -- not a broken or unwanted one. And the two refresh emails that would re-activate it (124/125) have never been sent (still drafts since 15 Jun). The channel isn't failing; it's switched off.
Source note
PostHog events
Referral opened/sent/signup/joined/activated.joined/activatedonly began firing May 2026 -- no paid-conversion history yet. Reward/payout data lives in the Customer.io "Referral Rewards" custom object, not PostHog.
Because the program is one-sided with negligible fulfilment overhead, referral CAC is approximately the reward. Here's LTV:CAC across a range of LTV (Jamie's April/March figures were $3,723 / $4,120 -- treated here as the mid case, to be confirmed against finance/Stripe):
| LTV / Reward | $300 | $1,000 | $1,500 | $2,000 |
|---|---|---|---|---|
| $2,500 (conservative) | 8.3:1 | 2.5:1 | 1.7:1 | 1.25:1 |
| $3,000 | 10:1 | 3.0:1 | 2.0:1 | 1.5:1 |
| $4,000 (Jamie's figs) | 13.3:1 | 4.0:1 | 2.7:1 | 2.0:1 |
Healthy-SaaS benchmark is 3:1. At $300 we're running a 13:1 channel -- wildly efficient, which is exactly why it's the wrong setting: great unit economics, near-zero volume. We're leaving the channel on the table to protect a ratio nobody needs. At $1,000 we sit at ~4:1 on Jamie's LTV -- still healthy, and roughly on par with our paid CAC ($600 to $1,000; April US $900) but with a higher-quality, pre-validated customer (3 months paid before we pay out). At $2,000 we need the RCM/eRx attach to clear 3:1. $1,000 is the defensible first move; $2,000 is a follow-on once volume + attach are proven.
US accounts carry structurally higher LTV: higher flat fee + RCM (3.9% of billings) + eRx attach. Modelled transparently (assumptions, not facts -- need finance validation):
This is the real argument for a US-specific $1,000 (and the reason Jamie's instinct to "push out a USD one" also simplifies operations): the US blended LTV plausibly sits at $5k to $7k, which carries $1,000 easily and arguably reaches toward $1,500 to $2,000 once attach is real.
A $1,000 reward on a non-RCM account is cash-negative in-year: by payout (~month 3) we've collected ~$135; subscription-only breakeven is ~22 months of tenure. It's a profitable lifetime investment, not an in-quarter one. The July campaign needs its own budget line, sized as acquisition spend. With RCM attach, payback compresses sharply -- another reason to point the high reward at the US.
The one-sided structure means today's gaming math is mild: to collect $300, a fraudster must fabricate an account and pay 3 months full price (~$117 to $147), leaving an arbitrage of approximately $150 to $180. Existing IP / email-domain / location matching + manual review handle that.
At $1,000 the arbitrage jumps to ~$850 to $880 per fake account (~5x). That crosses the line where industrialising fake accounts becomes worth someone's time. Mandatory with the increase:
Reframing the "$1 for 3 months" worry Carlos & Jamie raised: that exploit needs a referee-side discount to exist. The current program has none, so it doesn't apply today. It becomes live only if we later add a two-sided referee incentive -- at which point the 3-month gate must lengthen (Jamie's instinct was right). For the patient flip-side, where a referee discount is likely, the period must be extended from day one.
Keeping Phase 1 to a single variable is what lets us actually attribute the lift. The fraud guardrails aren't an "offer change" -- they're protection, and they ride along.
create staff runs ~1,000x/mo (10x via other paths) -- so replacing it costs almost nothing in lost invites, but for the same reason its referral upside is modest. Worth doing as a near-free placement; don't expect it to carry the channel. (Carlos already clocked this double-edge.)Competitive context (directional, not benchmarks to copy)
Heidi shows ~$50 but is a paid-ads-led model burning enormous CAC -- not referral-led. Customer.io's two-sided $2,000 + $250-for-a-qualified-demo program literally drove Carlos himself to refer 3 people -- behavioural proof that a big number + a low-threshold action moves real users. The $250-for-a-demo stage-gate is interesting but explicitly deferred -- don't stack it onto the reward increase (changes two things; some users find it cynical).
The advisor floated "why can't patients refer practitioners?" The numbers are tempting (Jamie: "388 today"), but I checked:
register runs ~300 to 430/day). Real client/patient creation is ~700 to 1,300/day (create client).Recommendation: defer behind the practitioner-side wins. If pursued later, start with a passive, un-incentivised portal placement + build patient-side telemetry first -- don't lead with an incentive or outbound contact. (Carlos & Jamie already leaned this way; the data backs "defer.")
The genuinely interesting long-game here is the B2C2B dynamic -- a client asking their practitioner to use a better system. Worth a real test eventually, but only after the core program is humming and measurable.
Both DRAFT, never sent, sender "Jamie from Carepatron":
| 124 -- admins | 125 -- staff / non-admins | |
|---|---|---|
| Audience | Paying + US/GB/CA/AU + admin = 3,358 | Paying + US/GB/CA/AU + non-admin = 3,472 |
| Reward in copy | flat $300, no cap, 3-mo-paid gate | same |
| CTA | app.carepatron.com/Referrals | same |
| Verdict | Ready to send | Ready, but check one thing |
Fix before sending:
Hold to decide: if Phase 1 ($1,000) is imminent, is sending $300 now still the right pre-step? Carlos's rationale (awareness + believability of the later bump) holds -- but it's a judgment call worth a beat. Nothing has been edited or sent; these are read-only findings.
create subscription + upgrade subscription combined).create staff ~1,000/mo.