My first kit was awful (a gift way back in the day), so I've been putting recipes together ever since. I split it into two categories: something to match a style (Doppelbock, Pilsner, Kolsch), or just trying to make an experimental yet very drinkable beer.
What worked for me when starting was to take something like a Fat Tire clone and just trying different things. I've also emailed most breweries in my area to ask how they put together specific brews (IPAs, ESBs) and for the most part they're extremely helpful.
IPAs are the test-bed. As a guideline I limit to 4 or fewer malts and limit crystal malt usage to no more than 5%. I go through phases, like weaning off of CaraPil and now using MO as a base malt and exploring how to get a maltier profile from that.
I rotate through various hops one by one to get a handle on each flavor and bitterness profile. It takes time. Each recipe is heavy on hop usage. I've been working through the bolder hops like Chinook, Apollo, Nugget, then Cascade and Centennial, then Perle, Willamette, Opal, Saaz, and so on, but trying to keep each recipe simple so the flavor profile is obvious.
I think in terms of n-factorial experimental design, so consider all aspects of input and figure out how to sample along the way. For example, splitting a batch of wort between 2 yeasts, bottling some when the yeast is finished, then dry-hopping the two batches 2 or 4 different ways, then bottling and kegging. If I want to go one step further, say trying bourboned-oak or perhaps an infusion (as in garlic + lemongrass + ginger), I'll use flip-top bottles and try it on a small portion of the batch.