The basic idea of geocoding is taking a human understandable representation of location (ex: 123 W Main St) and using a geocoding tool or service to transform that into a digitally understandable representation of location (ex: Latitude/Longitude coordinates). And with that basis, it can be seen that while an address is typically what geocoding is used for, it is by no means the only possible use as there are lots of ways people describe a location of various features other than lat/long.
For example, you may geocode a house using it's address (ex: 123 Main St, city, state, zip) which would put it part of the way down the correct main st, on a particular side of the road.
However, you may not have a full address, you may only have, let's say a list of customers or research participant's zip codes. Well, you could create a geocoder based on a point feature class of zip code centroids and run the dataset through there to quickly map out where all of the people in your dataset are from (a destination retailer may use this to understand what percentage of their customers are local and what percentage come from further away, how far people are willing to travel to come to their store, and what sections of a city they want to invest in putting up billboards in, or such decisions as that).
Similarly, you could take a point feature class/shapefile/geo-enabled dataset of business names (even if addresses are not available) and create a geocoder based on that so if someone calls 9-1-1 wanting to report an emergency at "XYZ Business", the dispatcher taking the emergency call can type in the business name and it tell them where the emergency is. After all, if you can help it, you don't want to have to say, excuse me, I know someone is bleeding to death right next to you, but could you walk out to the road and check for a mail box and tell me what the address is of the store you're in so I can get an ambulance to you. And equally you can set up geocoders that have road intersection information so you can geocode Main St & Center Rd instead of a specific address.
Furthermore, if the idea of a geocoder is to be able to automatically determine that 150 Main St is half way down the 100 block of a particular road, well then you could equally apply the same type of inquiry to a dataset that has nothing to do with a typical address. Lets say a oil pipeline company wants to locate all of their valves, meters, cleanouts, etc... along a pipeline. Well they could build a geocoder that basically says instead of this segment of road representing the 100 block of main st, this section of pipeline represents mile 1, this section is mile 2, etc... So when they have a facility to check on, have a leak, need to do repair work, etc... they can call in something like I'm just past mile marker 261 on pipeline ABC123 about to do a replacement of a valve. If you can geocode 123 Main St, there's no reason you couldn't create a geocoder for 456 Pipeline ABC.
And there are a whole host of other uses and styles of geocoders built by particular users/organizations for their particular needs and available data.
Hopefully that helps clear up some of the confusion that geocoders are only for addressing. It is true you are likely to only find commercially available pre-built geocoders or geocode services for addresses, but there a wide variety of other possible uses if you've got the right data to geocode against.