Start with "No."
The most expensive thing you can do as a startup is to add a product feature. I think most leaders have no idea how much a new product feature costs!
Calculating The Fully Loaded Cost
Engineering time itself is far more valuable than money - you cannot simply exchange money for good dev hours like a vending machine.
Each new feature adds ongoing costs to document, QA (with each release), support, and maintain (regression testing with every launch and work on it with every system upgrade, iOS update, etc.), which means you need to hire more engineers.
It makes your product more complex = harder for customers to use (and CS to support), which can diminish customer value and narrow your appeal to a handful of power-users.
It makes your proposition more complex, which makes it harder to market and sell. ( 'product marketing’ pulls marketers away from demand gen activities).
Therefore, when you’re deciding what to build, the default option should be “no.” And each new idea should have to fight its way onto the roadmap.
How to Decide?
Each new feature idea is based on an implicit (unproven) assumption that it will boost acquisition and/or customer delight. Make that assumption explicit and draft it as a hypothesis (if-then prediction with a measurable customer behaviour outcome). Find a way to validate that assumption before you build the feature. Here’s an amazing post from Ash Maurya with lean experimentation techniques that do not require building product.
Acknowledgements: Thanks to Mo Syed for helping me understand the true fully loaded cost of a new feature, and to Danny Yau and Robert Desmond for reviewing drafts of this missive.