Traffic can grow while contribution shrinks. Put an explicit cost budget around free trials before distributing a tool or expanding a search campaign.
Use one cohort, not unrelated dashboard totals
Define a cohort by trial start date and allow a fixed conversion window. Count all trial usage, including people who never activate, retries and failed calls. Record activation using a meaningful product outcome; a page view or SDK script load is not activation.
Trial acquisition cost per payer = total trial-cohort variable cost / converted payers
Contribution available after trial cost = observed paid contribution − trial-cohort variable cost
Paid contribution here already subtracts the selected fees, refunds and paid-period AI cost. It is not subscription revenue and should not be reduced by those costs twice. Keep currencies and periods aligned.
A concrete example
200 new trials cost $60 in total model and tool usage. Forty trials activate and ten become paying accounts within the defined window. Trial cost per payer is $6, while cost per activated trial is $1.50. If those ten payers contribute $100 after paid usage and fees, $40 remains after the trial cohort's cost. The other 190 trials are not free just because they did not convert.
If only five payers convert, trial cost per payer becomes $12. If zero convert, the denominator is zero: report $60 spent and no finite cost per payer. These are synthetic examples, not production results or a forecast.
Choose limits around useful activation
- Measure the minimum successful task that demonstrates value.
- Budget enough attempts for that task plus explicitly bounded retries.
- Keep expensive optional workflows separate from the initial activation path.
- Explain usage limits clearly; never silently downgrade success quality to hit a cost target.
- Compare cohorts after the same conversion window has elapsed.
A campaign with lower cost per click can still be worse if its users never complete the task. Use the agent outcome cost calculator before changing trial limits.
Connect trial budget to price
Use the existing AI pricing break-even calculator for per-plan usage allowance, then subtract the acquisition cohort's trial cost separately. A first-month result is not lifetime value; do not assume retention, upsells or future gross margin without observed evidence.
Missing prices or unobserved failed requests make a complete cohort cost unavailable. Label it partial and reconcile the gap. Do not scale a campaign simply because an incomplete dashboard displays a large margin.
What this evidence can and cannot prove
This is an arithmetic budgeting method, not a recommendation to change billing or trial limits automatically. No paid forecast or production customer data is used.