For an existing product, you have two primary modes of working on new features:
When optimizing an existing product to the needs of it's user base, or to extend into adjacent markets, research into real user behavior is critical. You would be hunting primarily for pain points, under-used features, and workflow gaps. You can guess based on tribal knowledge, but your risk would be much higher.
Even existing products have room for something totally new. Something your user base (maybe any user base) has no experience with and may not even know they need yet. You can and should do research, but someone with a talent for seeing the unseen is your best asset here.
In either case, be ready to ship often.
If the product team wants to get ahead of the users and start building to their assumptions, you have to also commit to rapid optimization and course correction. This is less of a concern when the project is heavily rooted in real user research -- still valuable, but not as critical.
Without research-driven feature definition:
- An optimization project probably won't be as successful as it should be. You'll have to watch the KPIs and pivot based on the post mortem.
- An innovation project will be a guess. Cross your fingers and hope that your product team is really good -- they just might be on to something.
I've worked on heavily researched optimization projects that yielded mediocre results. I've also worked on revolutionary innovation with little user validation that completely blew expectations out of the water. It all comes down to the talents of your product team, the resilience of your users, and a big lump of luck.