Start simple/evolve toward accuracy
With all that said to scare you away, why should you even bother forecasting? Because it helps you look at a particular type of decision in a consistent manner. Decisions will be made in a though-through manner with the proper inputs.
Also, people like to have a stake in the ground to say yes or no to. A forecast provides a good barometer to decide against. If you decide the forecast is wrong you will have some set of reasons to back up that conclusion. If you decide the model is correct, it has a whole basis of reasons to point to. If it is ultimately wrong, it is then easy to diagnose what happened to avoid the situation in the future.
For these reasons it is important to start any model simply. Starting simply allows complete understanding of how it works. Black box decision making processes are rarely effective. How can you possibly diagnose an issue with the process if no one understands how it is working?
Follow the “Keep it Simple, Stupid” methodology for the first few iterations. As it develops a track record begin adding on to it. Allow it to analyze exceptions, look at specific scenarios and take in more variables. Never add too many at once because if it breaks you’ll be back at square one with an undiagnosable problem and a wasted model.
Repeatedly test the model against past performance to see if it is more or less accurate in those situations. Always strive for improved prediction performance in the situations you’ve encountered before. If you move farther from the actual outcome the changes need to be tweaked. Past performance is as good an indicator of future performance as exists.