This is obviously a "one person's experience" account. I've done one deck. Basically it was done as fan service. I don't read the cards, myself, although I have 7 or 8 decks, which I bought for either the art or the concept. So I am not really a Tarotist. I'm a designer. I have no idea whether my deck would give a reader of Tarot cards a viable base for doing a reading from. But it is a fairly straightforward Rider-Waite homage and it is tied as close to the original interpretations (acto the old Eden Gray [Grey?] descriptions) as I could get it.
I suspect that such things are always a matter of metaphor.
Mine is a Harry Potter-themed deck. (Hence the fan service.) I'd noted that Tarot-themed fan art was a fairly popular continuing motif in any number of fan fests and gift exchanges, and one of my publication projects in development for eventual posting on my own site was such that using the Tarot as decorative elements for the cover and frontspiece seemed appropriate. At that point I had no intention of doing a whole deck.
It was nearly two years later that the project spun off into the challenge of building a full deck. My approach was to assign Potter characters to specific cards and to depict those characters as that card's subject. Fortunately, the Potterverse has a cast of dozens. There wasn't a shortage of potential subjects. Admittedly, the interpretation of a number of them and the metaphors used, hark back to the fanfics which the original cards were developed to enhance.
My approach was to examine the card image, and what Gray (and Stuart Kaplan who did the Hanson Roberts interpretations booklet) had to say about the symbolism used, and to compare it with what their interpretations of the meaning of that card was, and then to cast about for which Potter character would best fit that interpretation. I dismissed considerations of age and gender for everything but the Kings and Queens, but with only a couple of exceptions most of the cards subjects did still maintain the traditional gender.
Then, since I was working in low-end 3D software, I needed to adapt a model into what I hoped would be a recognizable interpretation of that character, dress it and build the general scene depicting that card's purpose in relation to a traditional deck. Since I was working in a program which is not a modeling program (and since I do not know how to actually create models of my own) I then needed to dress the scene as well as I could with available model props which either matched the symbolism traditionally used in that card, or something which would at least harmonize with it.
Some characters are depicted more than once. Some are depicted at different ages, but the intention was to try make them fit the traditional interpretations.
I did have, well, not a spreadsheet, but a chart with the various cards laid out and the characters penciled in according to their assignments. These assignments shifted about considerably as the project progressed. A few of them were still shifting right up to the last couple of weeks.
I still think that my biggest glitch was in not having considered actually getting the cards printed (for I originally intended it only for web posting) and finding a printer and downloading the template first. In fact, the online images do not include the modifications which were necessary to make the design fit the template. But upon the whole I am satisfied with the end result.