Draft · nothing built or live
Growth · for Callum
5 October 2026
Personal deadline test
A deadline of their own
Callum's idea: give each trial a short, expiring offer of their own, instead of waiting for the calendar month to end. My read: worth testing, but not for a big conversion lift. Deadlines clearly change when people pay. Most of the extra buyers they bring are people who pay on sign-up day, so a code late in the trial will probably add only 1 or 2 points. The real prize is matching today's conversion without a calendar-month sale.
34%
of trial payments land in the last 5 days of a calendar month. That's 5 days out of about 30.
268 of 799 US payers, sign-ups May–Aug 2026
15.1%vs 11.8%
30-day conversion when the month-end deadline falls inside the trial, compared with after it.
5,646 US sign-ups, May–Aug. Leave out sign-up-day buyers and it's 11.3% vs 10.1%, which could be chance.
46%
of US sign-ups (the 1st to the 14th of the month) reach the month-end offer deadline only after their trial ends.
2,610 of 5,646 US sign-ups, May–Aug
The evidence
Deadlines change when people buy. The calendar hands them out unevenly.
Someone who signs up on the 25th meets the month-end deadline on day 5 of their trial. Someone who signs up on the 3rd meets it on day 27, after the trial has ended. A personal deadline gives everyone one at the same point in their trial.
When people pay
Payers per day of trial, US sign-ups who paid within 30 days
Show data
| Trial day | Payers | Per day | In month's last 5 days |
|---|
The sign-up day is the biggest single day (24% of payers). After that, the last 3 days of the trial are the busiest stretch: a deadline people already have. Today's trial-end email offers $1 for the first month with code DOLLAR1, which has no end date.
A deadline inside the trial lifts conversion
30-day conversion of US sign-ups by the date they signed up. Small print: the trial day the month-end deadline lands on.
Paid during the trial (days 0–13)Paid after the trial (days 14–29)
Show data
| Signed up | Sign-ups | During trial | After trial | 30-day total |
|---|
Whoever gets the deadline inside the trial buys inside the trial. Early-month sign-ups partly catch up when the month-end arrives. But most of the 3.3-point gap is sign-up-day buyers (4.3% vs 1.9%), many of whom probably signed up because of the offer. Leave them out and it's 11.3% vs 10.1%, small enough to be chance. A code sent late in the trial reaches that second group.
What I'd change
Keep the idea. Tie it to the trial, give it two days, and make it real.
1
Tie the deadline to the trial end
Every trial already ends on a known date, and the last days of the trial are already the busiest stretch after day 0. Swap today's trial-end offer for the personal one, and make it end with the trial. That's honest, needs no new trigger, and changes one email we already send.
A trigger on the second session is a good second arm once the plumbing works. It lands earlier, closer to day 0, where buyers cancel within 30 days more often (10.1% vs 6.6%).
Callum's options: around day 10–14, or when they come back for a second session. This is the first.
2
Two days, not 24 hours
Open the offer on the morning of day 13 and close it at 11:59pm on day 14, in the customer's time zone. Practitioners are with clients all day, so a code that arrives at 9am and dies 24 hours later can be missed entirely.
The last day does the work either way. In offer months the final day is always the biggest (53 code starts on 31 Aug, 55 on 30 Sep). Two days gives us one launch and one last-day reminder.
Callum's version: 24 hours. This is about 39.
3
Make the deadline real, and big enough to read
The code is single-use, tied to their account, and actually expires. Then the test group gets no other offer messages until day 45: no calendar emails, no $1 win-back from campaign 73. If another offer turns up a week later, the deadline was never real.
Split new US trials 50/50. A small cohort can only tell us if something breaks, not whether it converts better (see the sample sizes below).
Agreed in Slack: exclude the test group from the calendar-month promos.
What the customer sees
Two emails, one countdown, one text, then it ends.
A clickable mock for one test-group trial that started on Tue 6 Oct. The countdown is live. Copy is a first draft; no prices are shown because the annual list prices still need checking.
Email
Carepatron · Unsubscribe · Privacy
In the app
app.carepatron.com/calendar
How the test runs
Half of new US trials keep today's journey. Half get a deadline of their own.
Who
New US trials
Practice owners, spam-screened, about 1,650 a month (1,628 in September). Split at random when the trial starts.
Left out: the hand-run lifecycle cohort. Everyone else who is split counts in the read, including people who pay before day 13 and never see the offer.
Control: today's journey50%
- Calendar-month offer emails, in-apps and texts
- Trial-end email: $1 first month (DOLLAR1, no end date)
- Post-trial $1 for 3 months (campaign 73)
Test: personal deadline50%
- No calendar-month offer messages
- Day 13: email + in-app countdown with their own code
- Day 14: last-day email, and a text if they've consented
- Code expires at 11:59pm on day 14; no other offers until day 45
The question is "no worse without the calendar sale", not "5 points better". Read each person at day 45. Both groups still see the website banner and pricing-page offer, because those are site-wide. That makes the test conservative: if the test group holds up anyway, the calendar promos weren't doing the work.
What we measure
| Paid within 30 days of sign-up Primary | Same definition as Callum's forecast. US baseline is about 14% (13.6% for May–Aug sign-ups). Pass = the test group is no more than 3 points below control. |
| Cash collected per sign-up, first 90 days | An annual plan collects a year up front; $1 first month collects $1. Conversion alone would undersell the annual offer. |
| Paid on days 31–45 | Did the deadline lose people who would have bought later, or only bring purchases forward? |
| Refunds, cancels and active days by day 60 Guardrail | Callum's point: people buy at peak excitement, then stop using it. Annual locks the revenue but can leave unhappy customers. |
| Unsubscribes Guardrail | The test adds a last-day email and text. |
How long it takes
Weeks of sign-ups at a 50/50 split, before the 45-day read. Estimates: 80% power, 5% significance, 14% baseline.
| What we can show | Per group | Weeks |
|---|---|---|
| No worse than −3 points | ~1,650 | ~9 |
| No worse than −2 points | ~3,700 | ~20 |
| Better by +3 points | ~2,280 | ~12 |
| Better by +2 points | ~5,000 | ~26 |
| Better by +1 point (best guess at the real size) | ~19,500 | 100+ |
We can't afford to prove a 1-point lift. We can show the test group is no worse in about 9 weeks, plus whether it collects more cash up front. The downside is small: 2 points worse for 9 weeks costs about 35 subscriptions in total, under 10% of one month (estimate).
What it takes to go live
None of this is built. It's mostly configuration.
Customer.io: the split and the messages
A random 50/50 split when a US trial starts, saved on the profile. The test group is excluded from offer broadcasts, campaign 1's trial-end offer email and campaign 73. Two emails, one in-app and one text, built as drafts.
Carlos
Stripe: one coupon, one code per person
A 50%-off-the-first-year coupon, with a single-use promotion code per person that's tied to their Stripe customer and expires at the end of day 14. Stripe promotion codes support all three. Still to check: how to limit it to annual plans. If Carlos builds it, he needs Stripe write access.
Julian or Carlos
The glue: make the code, put it on the profile
When someone enters the test group, an n8n step creates their code and writes the code and its end time to their Customer.io profile. Emails link to checkout with promoCode=, the same way the October offer links work.
Carlos, Julian
Measurement
Customer.io already records which promo code each new subscription used (automations 106 and 107). Add the group to PostHog as a person property so the day-45 read is one query.
Carlos, Callum
Decisions for Callum
Three calls before anything gets built.
Is the offer really better than the public one?
"Just for you" and "goes back to 6 months" are only true if 50% off a year isn't public that month. October's BREATHEANNUAL is exactly 50% off 12 months, on the site banner.
My call: start in a month with no public annual offer, or make the personal offer different (for example a longer annual discount).
When do we start?
October has BREATHE and the hand-run lifecycle cohort. November likely has Black Friday, which the test group would skip. That's the real question ("can we leave the calendar?"), but it's the hardest month to win.
My call: run across two month-ends so one promo's quirks don't decide it. Set the start once November's offer is known.
How big is the first group?
A full 50/50 split shows "no worse than 3 points" in about 9 weeks of sign-ups. A small group can't read conversion at all, but it does catch broken codes, links and exclusions.
My call: a 20% pilot for two weeks to prove the plumbing, then 50/50.