Step 1. Identify dependencies. Nothing will cause a project to go haywire faster than overlooked dependencies. There are two types: (1) internal dependencies that your team is responsible for and (2) external dependencies that you rely on other teams to complete. Know which is which and track both in your roadmap.
Don’t assume things will land on time. Regularly ask for estimated completion dates and pay attention if these dates get consistently pushed back.
Step 2. Estimate impact. The next step is to estimate the impact of each feature on your single success metric (e.g. sends). You'll want to synthesize all insight at hand, including prior experiment results, research, and analyses. Don't worry too much about being precise at first — your guesses will most likely widely miss the mark. What matters more is the relative ranking among the features you're considering.
Step 3. Record confidence. Ideally we are invested in a diverse portfolio of features: some sure things and other ambitious big, high-beta bets. What separates these two, outside of the effort involved, is our confidence. After recording your expected impacts, share your confidence in each feature. Then multiply these two numbers together to get the expected value of your product work.
Step 4. Estimate effort. In the final step we’ll estimate the effort involved. Work with your engineering manager or engineers to estimate how long it will take to build each feature. With one final calculation, we can get the expected ROI for each feature. Take the expected value of each feature (impact * confidence) and divide it by the cost (engineering effort). We now have all the information we need to prioritize.
Build the feature that have the highest ROI and no blocking dependencies. Or, work on the dependencies that unblock a high ROI bet. Putting this all together, we can create a table that spits out the highest ROI projects.