More console ui bugs related to pricing changes
The pencil banner in my free account is broken. The badge component is unreadable and I’m not sure if this was intended by both CTAs show the same billing table when i was expecting to see the usage on this account that led to the overage.
7 Replies
fair-roseOP•2w ago
FWIW guys, I’ve been using Neon for personal/hobby and company projects for over two years (the project in the screenshot was created Aug 2023) and this pricing change rollout has been jarring. I’ve never experienced an infrastructure company roll out such a major change in a way that (to the outsider at least) seems so haphazard. It’s disappointing to think that because it probably only negatively impacts affected free projects, it didn’t matter as much.
conscious-sapphire•2w ago
Thanks for reporting. On-call is already on it! 🙏
Hey @bertrand, any feedback would be much appreciated. What makes it seem haphazard to you?
For more context: Neon serves hundreds of thousands of free tier users and projects and only a tiny amount has been be affected/forced to change usage due to the changes. For those users, we have proactively upped free tier limits (so their db doesn't shut down) and helped others that we didn't predict to hit the limits manually after they reached out on Discord/X. But this has also been fairly limited (a few dozen users as you can see here on Discord).
We already got some great feedback:
- Better heads-up for free tier users (point taken!)
- Pricing calculator would be nice
- Bring down cost for low-compute 24/7 h always-on use case
and are working on addressing these points 🙏
Btw, I sent you a DM on Discord a few days ago offering to bump your limits after I saw your message in another thread 🙏
fair-roseOP•2w ago
Well...
Context: I detected that a db that has been quietly* running for exactly two years was down on Aug 23 with an API error indicating some kind of usage limit has been reached and all access has been suspended...taking the service offline.
Notice period (for a PaaS): I don't know if the notice email(s) went to spam but I looked on the blog and there were a series of posts from Mike Jerome dated August 14/15 announcing the pricing change with the note "If you're reading this, you're already on the new Free plan (or will be within a few hours)." For a PaaS, that's a rather shocking approach to a pricing rollout. Even PlanetScale-whose approach was criticized-gave developers a month notice. I can only assume I missed an email last month but given the blog timing itself, i have reasons to doubt that. I could be wrong.
Scale of change: Exacerbating matters is that the change was pretty significant. Going from 191.9 compute hours to 50 is a decrease of 74%. Again, you guys are a PaaS, not just SaaS. Architectural decisions were made on previous promises and migrations can be complex, non-trivial tasks that require significant planning and effort (even for small DBs). The old official docs explicitly boasted that the free compute allowance was "enough to run a primary 0.25 CU compute 24/7". The new promise is that it's "enough to stay online during business hours, Monday to Friday". A funny but significant change in ideal use guidance. This is even more jarring given that last October, in the aftermath of a few high-profile reversals on free tier, Nikita wrote a blog post doubling down on the free tier and claiming that free tiers are financially viable for Neon. The blog was titled "10x-ing our Free Plan", claimed the unit economics had been so well considered you could 10x your free plan project limit and referenced the manifesto which extolled that Neon was "designed for cost". My point, I guess, is that this wasn't just a significant practical change but it feels like a significant change to the brand promise as well...that was changed seemingly overnight in a kind of rug pull...and then rolled out like it was just a patch-level change. I'm sure you guys considered this deeply internally...but i'm not sure where that consideration was communicated.
Design of the DX: Even ignoring the above, the technical execution is also suspect. The free tier is limited to a 1 day retention for monitoring so it is (seemingly) impossible tor me to get a picture of the usage that contributed to me going over the new limit. The few places where I was expecting to get reporting were apparently affected by a bug so all i saw was that i had used 0/50 compute hours. This bug persisted for over 10 days after the pricing change and ostensibly a change that triggered complete outages (it's a hard cap after all). That bug was then fixed and I was greeted by that, let's say less than useful, announcement banner that was obviously rushed out and was again pretty confusing.
I'll stop there since I don't mean to drone on. I know this stuff is hard and complex and things in the db space are highly dynamic at the moment. To directly answer your question, what made it feel haphazard to me was that a significant change — architecturally but also at the heart of the brand — was made with little notice or grace period and poorly supported by even the main admin UI.
Odds and ends:
- Releasing a pricing change in mid August where people in NA and Europe are likely to be on vacation is an interesting move. I'm not sure there's ever a great time but there are bad times (e.g. popular holiday periods)
- Are there docs on how to audit usage on free tier? I can't figure this out in the UI to save my life.
- A clear way to export your data if your db is suspended would make it feel less like you're holding it hostage and using that as leverage.
---
* except for the yearly instability periods...this year it was may/june
funny positioning change for a "modern db" to go from thinking about devs building online services to those building internal support systems...has your usage patterns been dominated by retool users?
I didn’t see your DM and I appreciate the gesture.
I also appreciate that the total affected group was small. It always is. That’s the nature of issues on platforms. It’s always those users. I guess it’s telling that y’all are taking that as a point of relief. I’d just contend that sometimes that’s the actual signal.
In any case, I really didn’t want to have to open up discord, wait for it to update, join a new server…all to just “complain” about an issue. It would have been much faster and easier to just stay annoyed, put my credit card in and make a mental note to migrate off when i have time. In fact, it’s exactly what i eventually did. I just felt I’d also be useful and share my experience in case it’s useful to someone at Neon. Totally aware I’m not the ICP anymore…i like to think i was at some point in previous years. And for over 3 years (2022!…time flies), the product y’all built served me well.
conscious-sapphire•2w ago
Really appreciate the thoughtful and constructive feedback. I'll share your message as is internally with the team. Totally understand your points and I think we could have done better in many areas.
fair-roseOP•2w ago
I don’t know why this occurs to me now but I loved how Railway killed its free plan. It felt empathetic and focused on a sustainable transition.
conscious-sapphire•2w ago
I'll also make sure to follow up on your points soon 🙏
Big difference though is that we did not kill the free plan and we're not planning to!
fair-roseOP•2w ago
I know. It’s just funny to me that I felt better about that experience. Still use them at scale to this day. I guess the monetary bit didn’t bother me at all (new paid price plans look very reasonable too).