Hi - thanks for invite. There are many NoSQL databases on offer, so perhaps the best way to approach your query is to look at CouchDB in the context of similar databases.
CouchDB is a (JSON) document store in the same class as CouchBase (which started life as a version of CouchDB combined with Membase) and MongoDB. There are other lesser known ones like RavenDB, Cloudant (another derivative of CouchDB), PouchDB, etc.
What I like about CouchBase is its http-based REST interface through which you can easily create and manage documents. The idea of a 'database for the web' that can be managed through http is interesting (even seductive) but it is worth noting that competitors like MongoDB do not entirely share this vision. If you want a 'cheap' and easy to use document store then CouchDB is great.
I'm afraid I can't find much more else to say I like about it compared to other offerings.
With CouchDB (and derivatives like CouchBase and Cloudant) you literally have a 'lake' into which to dump documents. You then have to write your own (map-reduce) processes to analyse the data. There is no query language (CouchBase have recently released a query language for their DB). Great if you have good Javascript and plenty of time.
However, coming from a relational DB background, I find the MongoDB offering far more enticing. Its data store is logically organised into databases inside which are collections (tables) The security model is familiar and strong - with the ability to secure objects based on users and roles. The query language is far superior and their aggregation framework means there is no need to create your own map-reduce functions. Adding a node to a cluster is quite easy although the learning curve increases somewhat for sharding. Replication within cluster nodes is automatic. Besides there are a good number of drivers in different languages for MongoDB.
ps: I'm not a MongoDB employee :-)