Trace · business · August 2026
214
Subscribers past day 90
+38 on last month
The only number that matters. Roughly 60% of supplement subscription cancellations happen in months two and three, and only about 38% of first-time buyers ever reorder. Everything upstream of this is vanity.
Every figure on this page is illustrative
What is real and what is not →There is no trading history yet, so these are placeholders showing what the screen would display once there is. The screens, the engine and the logic behind them are real.
Active subscribers
412
+9.2% on last month
MRR
£19,364
+11.4% on last month
Cost to get a customer
£23
-18.6% on last month
Day-90 retention
52%
+4% on last month
Do customers stay? By where they came from
% still subscribed
Best channel by a distance. Arrives warm, stays.
Cheap and steady. Scales slowly but does not churn like paid.
Payback is 14 months against a 44% survival rate. Not viable at this CAC.
Read this one carefully
Paid social costs 13× more to acquire and retains 27 points worse at month five. At £62 CAC against £22.70 gross margin, payback lands around month 14 while fewer than half survive to month five. It is not a growth channel at this price, and scaling it would consume the raise.
Where £47 goes
£22.7 gross margin at 48%. Costs of £24.30 include the partner share, which is why partner CAC is effectively a variable cost rather than a marketing spend, because it only ever gets paid on revenue that exists.
Where people give up
58% complete
Already taking loses 17 points on its own, more than the next four steps combined. Biggest single drop. Nine options is too many to scan. Worth testing a shorter list with a “something else” escape.
Are we selling safely?
How often we refuse to sell
11.4%
usual 9.8%
Waiting on a nutritionist
3
1 past SLA
Cases the nutritionist changed
18%
nutritionist changed the blend
Unsafe recommendations sent
0
ever, all time
Why we stop and hand to a human
We cannot sell yet, and here is why
No nutritionist has signed off the recommendations.
The 21 rules that decide what goes in someone’s blend are written but unsigned. Until a registered nutritionist puts their name to them, the software physically refuses to give anyone a recommendation. It is a hard stop in the code, not a policy someone could forget or wave through, and there is an automated test proving the stop works.
Hiring that nutritionist is the only thing standing between us and launch.
Separately: we are refusing to sell to 11.4% of people this month because of medication, pregnancy or age, against a usual 9.8%. Roughly one in nine rather than one in ten. Either a rule is catching more people than intended, or the kind of person arriving has shifted. Worth knowing which.
See the rules awaiting sign-offFigures are illustrative. Analytics are cookieless and carry no health data. Events record question identifiers and outcomes, never answer values. Staff access to any customer’s health record requires an elevated role and a logged reason.