SaaS Is Broken: Why Bring Your Own Cloud (BYOC) Is the Future
16 points by rahul1990gupta
16 points by rahul1990gupta
This makes some good points, particularly around security, and I think Bring-Your-Own-Cloud will get more popular.
But that being said, I doubt it’ll save much in the way of costs. Ingestion-based pricing might be the chosen metric, but importantly, does not reflect all their actual costs of running a business. Inflated bandwidth/storage charges also cover for other biz costs that aren’t directly billed.
If the SaaS gets hosted on your own cloud, SaaS companies will just change how they charge you. Even if the SaaS no longer pays for storage/bandwidth/etc, they probably have a higher support burden with BYOC. BYOC means more heterogeneous environments, older versions hanging around, increased support friction/delay, etc. On top of which, they’d have to figure out another way to cover all their indirect costs.
Don’t get me wrong, I think BYOC is pretty underserved. And I’m sure there are scenarios where people see savings, but overall, I don’t think costs will change much.
Having worked on a SaaS/“On Premise” product before, the support burden of BYOC is immense for anything non-trivial. It’s always a giant investigation to even figure out if the hardware you’ve been given meets minimum specs, let alone trying to get telemetry back out to debug issues with.
Agreed. Companies can fat finger delete or otherwise accidentally misconfigure something that requires the saas company to spend a bunch of time investigating with little information beyond “it’s not working”. From the outside it’s impossible to tell immediately if it’s a product bug or someone on their team doing something they shouldn’t have done. Either way saas is stuck with the time investment.
Further, the company I work for provides a BYOC on top of our normal paas offering and while we had many companies expressing interest in BYO before we built it, when it came time to pull the trigger, many passed on the offering. To this day we only have about 1% of our customers using BYO.
agreed. SaaS got away with raising profits considerably, they will not go easily into making less. Otherwise I doubt the business people in charge care one iota how it’s delivered(your cloud or theirs).
One counter-example is 37 Signals. Their BYOC offering is called Once, and they charge a one-time fee for a lifetime license. Their Slack competitor Campfire only costs $299, regardless of how many people you have chatting.
Campfire: “Free updates to any 1.x version” So they release v2 and suddenly it’s not a one-time fee anymore. How is this different than the previous iteration where companies rented you a license to run X version?
They also have SaaS offerings, Hey and Basecamp.
Shifting away from “SaaS” to “BYOC” won’t really change pricing in the long run. Almost every product and service out there is priced based on what the customer is willing to pay, not how much it actually costs to run.
GitLab charges the same for a license for self-hosting as they do for hosting it for you, because you’ll most likely be willing to pay the same for both.
This, 100% this. I work in this world and based on discussions with other engineers, I get the sense that engineers think prices are calculated based on some nice concrete rules and mathematics. The reality is so far from that. Pricing calculations are almost vibe-based, taking into account the size of the customer, their revenue, what solutions they already use. And the above comment makes some great points about op-ex costs shifting rather than going away entirely. It reminds me of the engineering notion of where certain decisions cause complexity to move rather than be eliminated entirely. It’s the same thing, that cost doesn’t disappear, it just gets re-allocated to other things, or on the sales angle, “spun” differently.
This is just how prices work: it’s basic economics. The only thing that makes prices approach costs is competition, which in turn is enabled by commodification. If you’re not comparing between different providers, then there is no competition.
A VM is a commodity: it’s standardized to the point where it doesn’t matter much where it’s running, so you can shop around to some extent. A service is not: the cost of you, the customer, switching to something else is something that the service is incentivized to keep as high as possible. They want you locked in, so they can charge whatever they estimate you can afford.
I’ve been at several companies that had both SaaS and non-SaaS offerings. They always went full SaaS.
The reason wasn’t strictly “we can charge more”, the reason was “these non-SaaS customers require 10x the support and thus are a huge distraction to the business.”
Most businesses don’t understand what cloud ownership means and will not put up the resources. They expect the same experience, and they often expect you to fix their services and setups. Hell is other people’s clouds. Unless this changes drastically I don’t expect SaaS is going anywhere soon.
Security is much more important to us than price when we don’t buy SaaS: we need to store our data in a physical place we control, away from the Internet.
We’re probably more vulnerable to a skilled, dedicated attacker targeting us specifically, but on the other hand we’re not vulnerable at all to attacks on our vendors’ Interner-exposed services. It’s a good trade-off, IMHO.
And as a bonus, we can refuse to install updates that mess up a product we use. Of course we’ll probably need to find a replacement, but at least we can do so on our timeline and not the vendor’s.
Personally, I’m observing some trends:
Companies I work for proudly decide that they’ll adopt SaaS for anything outside their core competencies, because economies of scale mean SaaS providers can do it better. And in many cases this goes well, but in many cases, they find out pretty quickly it’s much more expensive and there’s lot of activity to save costs, migrate off SaaS, etc. I have no idea if the answer is SaaS or no SaaS! But there’s a huge waste because I feel there’s not a good framework to pick those decisions. In economic terms, as mentioned in another comment in this story “almost every product and service out there is priced based on what the customer is willing to pay, not how much it actually costs to run”, which IMHO means that successful SaaS companies achieve very nice margins, but the savings are not passed onto their customers!
Some SaaS are making it easier to use your own compute in interesting ways. I’m thinking mostly GitHub Runners, but also Gitpod Flex. The SaaS provides a very easy to install package that you run in your hardware that just needs an outbound connection to the SaaS; no messing with firewalls, etc. The core of the SaaS is still out of your hands, so you’re still locked in, but you get a good chunk of the cost benefits of self-hosting. This approach seems extremely smart to me, but we have yet to see how this works out. (E.g. GitHub Workspaces has not followed the path of GitHub Runners yet…)
I find it unfortunate that they do not mention cloud stack. There’s also nice things like digital rebar and other types of amazon-like clouds. You can build yourself. Shape blue is a great vendor. Not that I am advocating, I just would have liked to have seen more. How do I do it myself? Options.
BYOC allows customers to run SaaS applications using their own cloud infrastructure and resources rather than relying on a third-party vendor’s infrastructure.
BYOC definitely has its merits, but I don’t see how this point is true, considering that BYOC typically means “run the data plane of your SaaS product in my own VPC on AWS/GCP/Azure”. The benefits of the model lie in data sovereignty aspects, cost management (e.g. can use spend pre-committed with your cloud provider), etc., but not in avoiding third-party infrastructure dependencies.