Y Combinator Interview, Reinterview & Rejection
by Lisa GumermanHi Lisa and Josiah,
Unfortunately, we've decided not to fund Sunshower.io this batch.
This was a tough call - that's why we asked to come back to interview a second time - because we really liked you two as founders. You're a team we would like to see in YC.
However, we weren't convinced that the Sunshower idea was promising enough to fund. AWS cost optimizers is actually one of the most common ideas that applies to YC, so we've read dozens of applications every year from companies working on this. That doesn't necessarily mean it's a bad idea to work on, but it probably means that there are some significant reasons why it's difficult to do, and we didn't hear any specific insights about why you wouldn't run into the same issue that other companies have. The biggest issues, as I understand them, are that actually implementing these recommendations is challenging without causing downtime, and that engineering resources are super expensive and scarce, so that spending engineering time to save money on AWS costs is a bad trade unless the cost savings are huge. In your case, it seemed like you had already run into a lot of these issues with your already customers, along with the issue of reserved instances.
That said, we assume that someone will eventually crack this, and maybe it will be you. It's possible that the rise of kubernetes and containerization will make these problems finally tractable. If you do make more progress getting customers and show that you can actually get them to take your recommendations and save them money, we'd certainly be happy to talk again next batch. And if you would be interested in brainstorming other ideas with me at some point, I'd be happy to do that too.
Thanks very much for coming out here to interview with us in person. I really enjoyed meeting you and hope you'll keep in touch.
Best,
Jared
-
6
Most just say, "Not interested" so this was actually a thoughtful response and maybe has saved you a decade of work
-
1
It was super helpful -- we'd been thinking of doing much of this anyway, so it just accelerates our timelines.
-
-
2
Wow, very specific feedback. Better than most VC's.
-
2
"...implementing these recommendations is challenging without causing downtime" (and in doing so) " is a bad trade unless the cost savings are huge"
Are there ways to accomplish the same task by looking at historical data from x months? I would imagine the data would be quite the same.
The feedback at least gives you a direction, problem and (because of "previous companies") proves that you're on the right path.
If I we you I'd go back to your customers and try to figure out a solution for them by working closely, maybe even offer at reduced cost while trying to figure it out. After all they were once a startup too.
-
1
We look at the last month's data by default, which we currently get from CloudWatch. The concern is that if you're not using a more modern tech stack, updating instance types can be a lot more than just changing one line of a script/file somewhere. We have three different approaches planned out, the most promising, long-term, probably being getting folks on Kubernetes and helping them manage it, including doing in-place real-time optimizations.
-
-
1
An entertaining story through the YC Process | https://danielhgranados.blogspot.com/2020/12/journey-to-center-of-yc.html
-
1
@lisa can I connect you with John, he's the VP of a company and I'm hoping he can use your service or just talk to you - so you can figure out a way to solve his problem
-
1
Sure thing! My email address is [email protected]
-
-
1
Thanks for posting this. I was also rejected, its nice that they sent you a reason.
-
1
That's one of the benefits of getting the reinterview, I think. We were hanging out with someone else who had interviewed on Tuesday, and he got a more rote rejection email, but had just done the one interview.
-
-
1
I am really interested in keeping hosting costs down, but my current setup is at just over $6/month and I've stress tested the setup at over 1,000x current peak traffic. So despite your site looking fantastic, my very frugality would keep me away.
So it seems like this product needs moderately price conscious users, like Hatchbox.io
-
1
We're targeting folks spending 200-200k/month. We're tweaking some pricing, though, so once we start charging, the first 3 instances will be free. That said, you're basically at every cloud provider's base cost at $6/mo.
-
2
This comment was deleted 6 years ago.
-
1
I meant 1000 times my current peak traffic (updated the comment). And yep!
Initially, I made a super minimal clone of the Phoenix Framework itself, built directly on Cowboy server. After learning a lot, I made a series out of that work and then installed the real Phoenix Framework.
-
-
-
1
How do you feel about this?
-
2
At first, pretty terrible. Now, actually quite good. I'd been having some of these concerns for a few months, and just hadn't quite admitted it to myself. We have quite a few ideas on how to MAKE the problem tractable, and the fact that they deliberated so long despite having seen the idea so many times feels good. We can execute pretty quickly, so assuming I can get MY act together on the sales front (ohhhh lawd 😂) I think we're in a really good spot.
-
2
Good luck on the sales stuff. @louisswiss runs a podcast on Sales for Founders, it might be useful to you!
-
1
Thanks for the shoutout @rosiesherry!
@sunshowerlisa - if you have any sales q's, shoot me an email (or make a post on IH and tag me so others can pitch in/learn as well). Happy to help!
-
-
-
