Trial and Error

One of the biggest difficulties and struggles I've found in 40k is finding tactics or units that reliably work.

When selecting what units should be used and where, I rely partially on mathematical analysis (Howling Banshees will cut down more Marines in close combat than Striking Scorpions, for example,) intuition (the above Banshees have fleet, which helps, but are very, very vulnerable to Heavy Bolters,) and fluff. Fluff and mathematical analysis are pretty clear cut¹. The intuition part makes up for the lacking of mathematical analysis, and intuition has to come from some where.

As an example, I've seen people poopoo a unit because it failed them miserably in the first game or two they used it. 40k has a huge chance element. Ignoring this when deciding which unit to use can be somewhat presumptive, as your dice might have been treating you badly (or whatever the luck phrase of choice is.)

My basic point is this: in order to truly grasp the effectiveness of a unit, it needs to be tried in different combinations and multiple times. Any other decision might be made in haste, ignoring potentially great tactics. This means you'll have to grind through many games in many situations, some familiar, some not, and pay close attention to how a unit performed.

Get cracking!

¹Okay, mathematical analysis can be as murky as you wish to make it. Operations research commands it's own topic of math, and I've yet to look into it and bring it to bear on 40k. Perhaps I should.