It's not a popular thing to say, but co-op games and solo games are practically the same. From a design point of view, that is. The only main difference is the multiple player game experience. Can you not play Pandemic with 4 characters by yourself? Can you not choose drawing another card on Friday along with your friend? So the difference is basically how you drive player experience. It is important to note, that this does not work for semi-coop games, since they are in their nature competitive. Also, leave out hidden traitor games where all players but one are cooperating. They are a different experience all together.
For a good solo or coop you need the players to win, but only sometimes. You need the players to win, but feel they earned it. How do you do that? Here are some things I came up with while designing solo / coop games:
Here we are, playing against a "bot", an evil game. The first important thing is: randomization. As much as I dislike randomization that can kill you on the spot in a competitive game, I love it in a coop game. We're in this boat together, we all lose or win together. If I'm playing a solo, that is also a good thing, I need the game to beat me in order to get me to want and play it again. I love controlling the chances and doing my best and still lose along with my friends or by myself.
Points in a coop are only a decoration. A real win condition is always great. First, I want to know did we win or lose? Did we end slavery in the US? Did we save the world from total annihilation? Then and only then, you may give me points or some other indication that will tell us how good we did. If quicker is better in your game, the amount of turns left can tell you how good you did. If a deck of cards that runs out or some track getting to 0 is an indication of losing, give me points on cards left or how far I am from the end of the track. The only reason to work on game points only is when you have a bot that makes points against you. But I personally don't like that option since it feels like a non-solo or coop game trying to be one.
If I arranged this post by order of importance, this upcoming one would be number 1. Well, maybe, but only since it is important in every game. Good decisions that matter. I wrote about this in another entry, but you should bare in mind that when there's one entity playing (one or more players) against a game, there must be something in that game that will make even 4 experienced gamers scratch the top of their hat racks.
Oh! One more. You need to have different strategies that will be viable in different situations. That depends on the randomization of the setup or a twist of events that took place. Unlike a competitive game, these strategies don't have to be 100% balanced for an average setup. It is OK to push players towards one strategy or another and make that one more viable for a certain game condition. Part of a game with a programmed competitor (i.e. the game) is finding its weak spots and attacking those.
In no way this is easy or harder than designing a competitive game, but you do encounter different problems when designing a player vs. bot game rather than a player vs. player game.
In the picture, a robot that may or may not be evil and also may or may not want you to lose at a game. Kidding, of course he's evil and wants you to lose. Badly.