Why there is no free plan, and a 14-day trial instead

A free SaaS plan and a free trial are both “free,” but they confirm different things. A free plan is a design that spreads continued use. A trial is a design that confirms value and willingness to pay within the deadline.

On our own SaaS, which does not yet have paying customers, I did not put a free plan in place. I limited it to a 14-day trial in which the range of the Standard plan can be tried. Card registration at the start is not required, and there is no automatic charge after the deadline. What I want to know first at this stage is not the number of registrations. It is whether anyone who has used it will pay the monthly fee.

This is not a conclusion that rejects freemium in general. A free plan may fit a service that can test a network effect, word of mouth from free use, and a clear paid upsell. This article shows what we prioritize at our company while customers are zero, and what we would need to learn before changing the decision.

I am building Shoryaku AI ( https://shoryaku.net ), a marketing-support SaaS, as an indie developer.

Current prices and the trial

At the time this article was published, there are two plans.

For 14 days after signup, the features can be tried without registering a card at the start. If no card is registered when the trial ends, use stops. It does not switch to a paid contract on its own.

The basis for extending the period from 7 days to 14 days is the public terms of 9 competitors, and the time we need to carry out a move ourselves and look at the numbers again. The research is written in Why so many SaaS free trials are 14 days: I checked 9 companies.

AI cost is not the reason

Explaining the lack of a free plan as “AI is expensive” does not match the current implementation.

In an estimate from the current code, the AI cost of the Standard plan is ¥319 a month at typical use, and the ceiling when the credits are used up is ¥1,238 a month. This is not the average of customers’ actual use. It is an estimate from the token amounts set in the implementation. The calculation conditions, and the correction of an older estimate of ¥140 a month at the time of publication, are summarized in Manage AI SaaS cost from both actual consumption and the monthly ceiling.

If free users are few, the cost alone is not at a level that immediately makes the product unviable. The problem is what can be learned when free use is increased.

With zero customers, measure willingness to pay before the number of users

With a free plan, registrations, continued use, and generation counts can be measured. Whether that person will pay ¥4,980 a month is unknown unless you design a path that reaches a paid feature, or an upgrade request.

We do not yet have the following numbers.

• The rate of reaching the first analysis from the start of the trial
• The share who felt the value within 14 days
• The conversion rate from the trial to paid
• Retention after 1 month and after 3 months
• The effect of a free plan on increasing referrals and content generation

Making a permanent free tier in this state increases the hypotheses that have to be measured. So at the start I made it one boundary: after showing features equivalent to Standard for 14 days, continuing requires a paid contract. It does not go as far as trying features that are limited to Pro.

When I looked at public listings on an M&A site, there were cases where the seller explained continued customers and profit as the value. Many of those were asking prices, and they are not data that proves free users are worthless. The current decision is not a general rule taken from an external market. It is to narrow what we want to verify first down to one thing.

Access limits are decided in one place, from the period and the contract state

In the implementation, the boundary between free and paid is not split by feature. A shared function looks at the contract state and the deadline.

The entrance of a feature that requires a contract passes through this decision. The feature difference between Standard and Pro is a separate permission check. Because the trial-deadline check is not written per feature, the range where a new entrance can forget to attach the deadline check gets smaller.

The trial deadline is the same time of day, 14 days after the registration time. To keep using it after the end, register a card from the management screen and subscribe to a paid plan.

Conditions for reconsidering a free plan

If any of the following can be measured, I will reconsider a free plan.

• Paid registrations increase from public content or referrals that free users created
• 14 days is not enough to confirm the value, and continued free use is easier to test than extending the period
• The difference between the free tier and the paid tier can be drawn clearly by usage volume, rather than by the number of features
• Even including support effort and AI cost, the acquisition effect of free use is larger

The other way around, I will not add one while the conversion rate and the referral effect are both unknown. The current conclusion is to decide the conditions for changing, and not to treat the absence of a free plan as the right answer itself.

Summary

• We limited it to a 14-day trial with no card registration and no automatic charge, rather than a free plan
• The reason is not that the AI cost is high. It is to test willingness to pay first, at the stage where customers are zero
• The Standard AI cost in the current estimate is ¥319 at typical use and ¥1,238 at the ceiling. The average of actual use is unmeasured
• Access limits were gathered into a shared decision on contract state and trial deadline, rather than per feature
• Paid conversion, retention, and referral effect are unmeasured. If the numbers come in, the free plan will be reconsidered

• Pricing
• SaaS
• Indie development