The work I am proudest of: one route that renders whatever the visitor has already done, whether chosen, claimed, shared, redeemed someone else’s offer or all of it, componentised well enough that the next campaign started from it.
The Halloween campaign asks for one decision: trick, an offer for you, or treat, an offer you can only give away, redeemable by anyone except the person who shared it. That decision is stored, which means the page is never the same page twice. Come back after claiming and you should see your offer and the option to go and share; come back after doing both and you should see a third screen; arrive on a friend’s link and you should see theirs. Getting it wrong means showing someone the chooser again for something they have already spent.
Every state and every way into it, including the ones only reachable by leaving and coming back. The list of screens fell out of that graph rather than out of the design file.
One entry point, not a URL per outcome. What is stored for that user decides which screen renders, so a returning visitor lands where they left off and no state can be reached by guessing a link.
A shared offer has two sides: the person who gave it, who must not be able to redeem it, and the person who received it, who gets a screen of their own with its own next step, and who can then be pulled into the other half of the campaign.
States, screens and the progress they read from are separate pieces rather than one page full of conditionals. San Valentín ran the same machine with a different skin and a different offer split, and the eme studios drop reused it for points-to-merch redemption.
Three campaigns ran on the same structure. The later ones were a reskin and an offer change rather than a rebuild, which is exactly what the first one was designed to buy.