The temptation with a new feature is to open the editor and start typing. I have learned to do the opposite, at least at the start.
On a recent build I was asked to add campaign and link tracking to a platform. Before I wrote a line of application code, I mapped the data. What are the pieces, how do they relate, and what has to stay unique so the whole thing does not fall apart later. I went through a few versions of that model before I was happy with it. Each version was cheap to change because it was still just a diagram, not a database full of real records.
Only after the shape was right did I break the work into small, ordered tickets and start building the typed API layer on top of it.
Getting the data model wrong is one of the most expensive mistakes you can make, because everything above it inherits the mess. An hour spent on a diagram has saved me days of untangling later. Slow at the start, fast the rest of the way. That trade is almost always worth it.