How do you decide what features should be free and what should be paid?

I'm building three products right now and each one has a completely different answer to this. For Shoute (voice-to-text), the free tier is generous because the product sells itself once people try it. For Fridge2Frame (physical keepsakes), free doesn't make sense since there's a cost-of-goods. My rule of thumb: if the free tier creates a habit, it's worth giving away.

As a new developer in the Crypto Tax space promoting Privacy - full disclosure is important to me. Therefor, providing users a privacy policy and access to all their calculations and reports is free. The only paid part of the app is printing final report PDF's and CSV's. They can use the software and manually transfer results to their tax submissions - and test results against their current methods. I am confident once they have all their data encrypted and classified, next year's taxes are a snap using my app. Plus, I am ultra competitive on price and security. For me accuracy matters too - users need to be 100% satisfied before they buy - because that's what I would want.

So - ask yourself the question first - what would I pay for this and where is the benefit? Then, be open with users.

This is what i consider when i want to determine pricing and free features.
If the components are already free in some products, i usually make them free too, and if i added new features on to them then the new features will be available as a paid option.
Another metric i use is the quantity usage of the feature, the more the usage increases, the more the user will have to pay.

I'm still new to this, so i'm open to recommendations.

I usually think about it like this: free should show the value, paid should deliver the deeper outcome.

Free features should let people understand what the product does and actually get a small win. If users can’t feel the value, they’ll never pay.

Paid features are usually the things that either save a lot of time, unlock more power, or scale the results. Basically the parts that become important once someone is already getting value.

For example:

• Free → try the core idea

• Paid → do more of it, faster, or with better results

I also try to avoid making the free version feel “crippled.” It should still be genuinely useful. If people like using it, upgrading becomes a natural step instead of a forced one.

It’s definitely a balancing act though sometimes you only figure it out after watching how users actually use the product.

One great idea I'm bouncing around at the moment is to offer paid subscription features for free but with limits. I believe the freemium model greatly increases conversions, and you can see examples of this in applications like Vercel / Expo, even Microsoft has a "only pay when you cross a limit" feature.

I think this works great, because it allows users to build up trust as you mentioned Nika, but it also increases the value proposition. I'd be more willing to pay for something personally that's already working for me.

This is a million-dollar question for any creator! Personally, I believe the 'Free' tier should offer enough value to solve a core problem, making the user trust your brand. The 'Paid' tier should then focus on scale, automation, or deeper expertise.

I'm curious, for those of you with SaaS products: Have you ever felt that you gave away 'too much' for free? How did you pivot from that? Would love to learn from your experiences!

one thing nobody here is mentioning is that the free vs paid calculus completely changes when your product has real per-user costs. i build AI-powered tools and every single API call costs money. image generation is like $0.003 per image, animation runs about 7 cents per clip, TTS has its own costs. so "free tier" doesnt just mean less revenue, it means actively losing money on every active user.

what worked for me: make the creative/exploratory part free (browsing, planning, basic editing) and gate the compute-heavy stuff behind usage limits. users get enough to feel the product, but the expensive operations are metered. went from burning cash on free users to actually having sustainable unit economics.

the trial vs freemium debate misses the point for AI products honestly. its not about time gates or feature gates, its about cost gates. let people do everything, just cap the volume. they self-select into paying when they hit real usage because by then theyve already built the habit.

I usually think about it this way: free should help people reach the “aha” moment, paid should help them get ongoing outcomes.

If the free tier solves the whole problem, there’s no real reason to upgrade. But if it’s too limited, users never experience enough value to care. The sweet spot is letting people feel the core value for free, then charging for depth, scale, speed, customization, or repeat use.

A simple test is:
Would I be okay if thousands of free users only used this part?
If yes, keep it free.
If no, and especially if it creates real cost or delivers compounding value, that’s usually a paid feature.

the thing nobody here is mentioning is how AI changes this calculus completely. when your most valuable features run on inference costs that scale per-use, the old "give away the core, charge for premium" model breaks down fast.

i build tools where every API call costs real money. image generation is like $0.003 per image, animation runs ~7 cents per clip, TTS has its own costs. so the free tier cant just be "unlimited basic" because basic still costs me money on every single request. usage-based pricing or hard caps on free tiers arent just a business strategy, theyre survival math for solo builders.

the counterintuitive thing ive found: being transparent about costs actually converts better than hiding them behind feature gates. telling users "this costs me X to run, heres what you get free, heres what the paid tier unlocks" builds way more trust than arbitrary feature walls. people respect honesty about unit economics more than they respect marketing frameworks about "wow moments" and conversion funnels.

something i rarely see discussed here is that the free/paid boundary should follow your actual cost curve, not some theoretical framework about user psychology. i run a content pipeline that uses multiple AI APIs and the costs vary wildly per feature. image generation is like $0.003 per image so giving that away for free is basically nothing. but video animation runs about 7 cents per 10 second clip and thats where it adds up fast at scale.

so my rule is simple: if the marginal cost of a feature is near zero, make it free and use it as the hook. if theres real per-use cost, thats your paid tier. not because of some conversion funnel logic but because your bank account literally demands it when youre self funded.

the whole "wow moment" framing is fine but it ignores that most solo builders arent choosing between free and paid. theyre choosing between sustainable and bankrupt. i went from spending $150/mo on one service to $40-50 by just caching outputs and checking similarity before regenerating anything. that kind of engineering is what actually lets you offer a generous free tier without bleeding money.