Pricing
Currently, I'm right on the edge of choosing Convex over Supabase. Have been reading the docs for a few days. However, the lack of clarity over pricing is holding me back. Especially related to the Function calls (25M) and Action compute(250 GB-hours) in the Pro plan. As far as I know, Supabase doesn't have such restrictions in their Pro plan (unlimited requests and shared CPU without any restriction on compute hours). Upon searching Discord, I couldn't find anything meaningful about it.
In fact, another user shared a similar concern a couple of days ago- "if you have 100 users in a chat room, and 100 messages are added, does it cost you 10,000 (GET) + 100 (POST) function calls usage?". Some users were discussing about it in August as well, when someone suggested a pricing structure like- https://aws.amazon.com/amplify/pricing/.
In fact, for my use case, Chatgpt gave me an sample estimate of 72M calls/month (for chat, subscription, updates etc) for a decently scaled project I'm planning. However, there is no certainity about it. It is just a rough estimate. Could be more, could be less. But if true, that estimate would probably add another $100($2/M) to the monthly bill. Had a nice experience with Convex so far and would like to build here, but don't want to be surprised with the 'serverless' bills later down the road. It would be a bit hard to change the setup then. So, it would be really helpful, if someone can provide more clarity about it.
Amazon Web Services, Inc.
AWS Amplify Pricing | Front-End Web & Mobile | Amazon Web Services
With AWS Amplify, you pay only for what you use with no minimum fees or mandatory service usage. Get started for free with the AWS Free Tier.
5 Replies
Thanks for posting in <#1088161997662724167>.
Reminder: If you have a Convex Pro account, use the Convex Dashboard to file support tickets.
- Provide context: What are you trying to achieve, what is the end-user interaction, what are you seeing? (full error message, command output, etc.)
- Use search.convex.dev to search Docs, Stack, and Discord all at once.
- Additionally, you can post your questions in the Convex Community's <#1228095053885476985> channel to receive a response from AI.
- Avoid tagging staff unless specifically instructed.
Thank you!
Yeah i get the confusion, i look at supabase and i feel like its even harder to analyze then convex pricing as they have overlap. There is not a lot of apples to apples comparisons. I think messages on supabase is "basically" the same as functions in convex so you'd be paying for a fee too $2.50/m. There is a lot of competition in this market so i assume at the end of the day the pricing is not going to be that far off. It would take a huge effort to provide a real detailed analysis.
We feel like the benefits with convex with transactions, etc and all dx to be worth whatever difference. And its not really worth our time until its 10's of thousands/month of dollars to roll our own non-serverless systems that would really make a price difference. Every hour a dev saves is the hardest thing to factor in. And its even more important the newer the company.
Willing to pay for their services(realtime, transactions, DX etc), but for a fair cost. It’s the surprise costs that could be annoying.
Here’s the summary of their Pro plans-
- Compute- 2-core ARM CPU / 1 GB RAM / Connections: Direct 60, Pooler 200
- 8 GB disk size per project (then $0.125 per GB), 250 GB bandwidth(then $0.09 per GB), 100 GB file storage(then $0.021 per GB)
- 5M realtime messages(then $2.50 per Million), 2 Million Edge Functions(then $2 per 1 Million)
Comparing it to Convex plan
- Action compute- 250GB hours(then $0.30 per GB), 25M Function Calls
- 50 GB database storage(then $0.20 per GB, 50 GB database bandwidth(then $0.20 per GB).
- 100 GB file storage(then $0.03 per GB), 50 GB file bandwidth(then $0.30 per GB)
- 1 GB Vector storage with 10 GB bandwidth. (then $0.10 per GB)
It’s mainly the Action compute(could be made lenient and easy for scaling) and Function calls(could be eliminated altogether) that bring the surprise factor here when compared to Supabase. Not sure whether Convex’s Function calls and Supabase’s Edge Functions are the same thing. But looks like they are not.
If Convex ultimately pays their own bills as per the compute used, it would be better if they charge simply on the compute with their own markup on top of it. That would simplify a lot of complications.
Aim should be to bring more clarity and determinism to the monthly bills without any big surprises. Unwarranted surprises while scaling up can be deal breaker for many. Convex could definitely be a better choice, if they pull this off neatly.
Also, if someone is reading this from the Convex team, kindly eliminate the 'hover over for more info' in the pricing table, simply introduce the details in the table itself to make it more transparent and readable. Not a good UX when you have to keep on hovering the mouse over them to get details.
a few things. and thanks for the feedback!
1. the pricing page is pretty old, and agree we need to revamp it. we're gonna do it when we launch the "scale" plan
2. we sort of copied vercel's conventions at the time, the whole hover-for-detail. I agree it obscures information. it's something we'll have to figure out for the next version. there's a bit of an information density thing to try to get right with a complex product like convex... especially on mobile
3. in practice, the pro plan ends up being fine for most users and they're pretty happy with their pricing. the one exception is if you truly have a TON of function calls and a high cache hit rate. the pricing is a bit unfair in that case b/c function calls simplifies underlying resource usage in a way that doesn't give you pricing benefits when you optimize. that's related to the scale plan as well... (most users, though, this doesn't affect b/c they actualy are more "db bandwidth" impacted than function call. again, it depends on how much. you serve cached data, R/W ratio, etc. we're always happy to talk it out)
4. the scale plan is the plan that will do more "pass through" resource pricing. it's just that price sheet is longer and more complicated. that plan tends to be better for teams who are really driving a lot of traffic. but the "platform fees" (i.e. seat fees) are also higher. the current pro plan is designed to be simple (some might say too simple), and "easy to buy", b/c most teams on it in practice are not exceeding their built-in resources. it's really for the phase in the company where you have no traction, and you want to model convex as a fixed $25/mo per dev (to agree with your "determinism" term)
the metered economics aren't really quite right for customers with true adoption. we currently solve all this manually on a customer by customer basis, but we want to codify what we've learned into the "scale plan"
5. in general, our incentive is to have you love the platform and build trust on us over 5-10+ years. so we have no interest in predatory pricing! finding the right price where it feels like the customer is getting a TON of value (and convex is keeping a meaningful minority of it, but not a majority) is the balancing act we'll always shoot for, b/c it feels win/win. pricing is just hard and takes time, tbh
anyway, quick brain dump on all things convex & pricing
the new plans will be something like "Starter/Launch (which is currently pro)/Scale" -- we like Neon's terms here
and we tend to find the pro plan is actually just fine, as long as the scale plan exists to be able to project future economics if/when your business succeeds 😄
feel free to fire away with any more questions
Thank you for your honest reply. Your 5th point sums up the expectations really well.
Some more suggestions upon my brief interaction with Convex:
1. It would be nice if you can relax the $25 Pro plan a bit more(compute and functions) with potentially no surprise bills as such 🙂
2. Neon’s pricing seems mostly compute hours based. Could bring more rationality in pricing. However, just keep scale(extra usage) pricing reasonable enough especially for indie/2-3 developers.
3. Branding and UI could be improved. Would be better if there was more information density with a minimal yet beautiful UI. Supabase definitely has an edge here. Default dark mode and matte tones can help improve a lot. And looks like Convex is already working on it.
4. A comprehensive video walkthrough series of the docs (maybe around 5-6 hour content) explaining the concepts while tinkering with the code in an editor would have saved time. It would have provided better context about Convex and made the transition even faster for new developers. Like this video- https://www.youtube.com/watch?v=WFyRL3igPkA but for the whole docs.
5. Lastly, better up to date integrated starter templates could reduce friction even more.
5. Lastly, better up to date integrated starter templates could reduce friction even more.
Convex
YouTube
Convex Masterclass: Writing Good Types, Part 1
Ian sat down in the studio to talk about his favorite tips and tricks when hacking on Convex projects in TypeScript. Grab some tea and chill out with Ian as he covers some potentially game-changing improvements to your development workflow.