Depends on how you’re measuring success. Are you looking for profit? Distribution? Both? Something else? Also depends on how you are funding development.
On premises is going to require support resources for installation and maintenance. You’re going to need a build & distribution team who cut a release and package it for distribution. You are going to have to develop to a release schedule cadence, you are not going to have continuous delivery & integration. You are going to have to pick and use components that can sustain that cadence. You are not going to want to use, for example, a library or framework that releases monthly, if you’re doing quarterly, bi–annual, or yearly releases.
None of this is wrong…it is just very different from how product development has shifted to the cloud.
A cloud SaaS shifts the resource allocation around. It doesn't necessarily cut it. You pick up continuous development/integration. You increase release velocity. You have a smaller build/deployment team but a larger operations/SRE team. If successful you become a critical component of your customer's operations. But, when successful, you also become a target for intentional information security attacks (depending on the value of your customer's information you’re storing and processing).
If your customer’s information is valuable or interesting to governments you become a target for legal and regulatory investigations. So you may need a bigger legal team if you’re providing a cloud based service vs on premises.
So…I don’t think there's a right answer. I think it depends on how you’re building the business, how you want to allocate capital, and how rapidly you want to build revenue.
Gut instinct is that a cloud based service will generate revenue and get you to break even earlier but an on premises solution may generate higher profits once you hit your stride.