What is happening now? What is reality saying? The amount of space that is looked at needs to be proportionate to the forecast time.
What patterns are seen? Was map analysis done? Are there certain areas that are colder than others? Are there places that have clouds and others that don't have clouds? What does the radar say? Why, why why why why? If you know the mechanisms that are generating weather now, then you can understand how they might evolve.
Before you use a weather model, you should understand it. Garbage in=garbage out, sometimes. How well did the model do today? If it is overpredicting temperature right now, will it continue overpredicting the temperature? Will the errors that occurred upstream yesterday occur today?
Taking a model at face value might work for research purposes (unless you are researching the model itself), but shouldn't be done on a practical level or in a rush. What does Model Output Statistics (MOS) say?
This is probably the step that requires the least amount of explanation. Though the more intricate the type of forecast, the harder and harder this becomes. Does it actually require numbers, or is there some sort of GIS software (like for hurricane trajectory or lightning forecast)?
This can't be stated enough. You must verify how well you did. Decisions need to be made on how the forecast will be verified. What data sources do you know of for verification? If I could move this up to number 1 and still have this make sense, I would. Because this is what you should start off with. This actually goes part and parcel with starting with observations, since observations are what you start a forecast with. Understand the processes of why your forecast was off. This will serve you in the future and in the next forecast.