How do you know when you have enough solid information to put forward a product suggestion?

You need to prove that there's market demand for it and that the customer is willing to pay for it. Or that it's a major cause of churn which clearly shows monetary impact. That can be validated in a number of ways: analyst validation, customer/prospect validation, willingness to pay research. A good place to start would be talking to Sales. Are we losing deals because we don't have that feature? What was the impact of that loss? What risk does it pose to the business by not building that capability? I would also talk to Customer Success (if you have that team). What are they hearing from customers? Are we seeing customers churn because we don't have that capability?

It's definitely should never be about just one thing - the customer wants it, our competitor has it, etc. This is a mistake that's often made when product teams build without including PMM, or other teams. There is that requirement to have multiple levels of evidence before a decision is made, however one I think is most overlooked is - ROI, mainly because it takes so much time and effort. I started my career as a financial analyst, and spent my days doing line by line forecasting of our product SKUs on a monthly basis to understand what our projected revenue forecast would be. I've tried to bring this discipline to my work as I suggest product features to be built. This additional information, whether or not we can expect return on investment by having it or not should be calculated, and I suggest PMMs try and work more with finance. It can be done based on current requests (win/loss analysis), growth rates, and understanding the type of customers that tend to buy, then working with the finance teams that are already doing the growth analysis and revenue prediction to identify how this new feature could impact based on the cost to build it. We are working on a similar project at Miro, looking at the overall cost from engineering support requirement, customer support, infrastructure costs, etc. Having this level of evidience along side your market research becomes invaluable. 

1. Focus group or survey data from customers or prospects
2. Market research study showing the need for certain capabilities
3. Survey data from sales, customer success showing the need
4. Deal loss analysis showing that lack of a capability is killing deals

If you have all of these, thats a perfect storm. But even one of these, provided its accurate and unique data can go a long way in convincing PM.

Ultimately I view roadmap as a path to your company's vision to get more customers to use your service. Thus, I definitely dont think that meet-comp features by themselves are justfiable evidence to get something into roadmap--because the market already has that and is it really going to differentiate your product. In certain cases it will help you not lose deals because of a gap, but its not really making you a visionary, so a healthy balance of when to add competitive features is nice.  

I'd say a must have feature is some combo of high forecasted customer usage, high addressable market opportunity and low cost to create. All of which map to what is the ROI behind building the feature? If you take that approach, and look across a few different features, you'll start to see opportunities that make sense for your business.