Practice will vary from place to place. I live in the middle of Ireland. I’ll try to describe general principles, but this may work out differently in different areas. Each congregation of Jehovah’s Witnesses has its own “territory” (roughly analogous to a parish in other denominations, perhaps). That territory will then be subdivided into maps.
The territory of my former congregation covered a few towns and a very large rural area. This is partly because of population demographics: Ireland tends to have a lot of one-off scattered houses in the countryside, while in the UK people are more likely to live in towns and villages. In Germany, I believe, even farmers tend to live in villages, and they commute to their farms in the morning. In Ireland, though, it’s quite common to have houses in the middle of nowhere which have nothing to do with a farm: people just like living in rural areas. Our maps, then, covered different types of territory. A person working such a map would be expected to cover the whole thing: houses and businesses alike. They would keep a record of people not at home, and attempt to call back at a different time of day.
Of course, some people prefer doing different types of work. Some of the urban maps had a note on them: “Businesses on this map are included on a separate business map.” A person working such a map would cover only the houses, leaving the businesses. (I was working on redividing the territory and creating new maps; the business maps were never actually prepared before I left the Witnesses. I assume they have some now.)
There are cases where things would work differently. The most obvious example is when congregations serving different language groups have overlapping territory. In this case, people might try to specifically seek out speakers of the language they are targeting. On one occasion, my mother (an ISL/English interpreter) and a couple of visiting Deaf Witnesses went out specifically looking for Deaf people.
Of course, not all witnessing activity takes place from door to door. In street witnessing, it is relatively easy (and fairly instinctive, actually) to pick out particular people who look more approachable, by whatever metric you use.
I suspect that more targeted witnessing happens in more populated areas, generally. In a rural area, once you’ve spent an hour travelling to the place, you’re going to try to get to everyone.
