Nearly every survey is different. For example, we've recently completed a two-day 'cordon survey' in Luton. This involved throwing an imaginary cordon around the whole of the town centre and monitoring every vehicle and pedestrian entering or leaving the cordoned area, in quarter of an hour intervals, between 7am and 7pm. For the purposes of that survey, it didn't matter whether vehicles turned left or right (or went straight ahead) when crossing the cordon, all that mattered was that we noted the type of vehicle and the number of occupants in the vehicle. (as stated, we also counted pedestrians). That survey was on behalf of a consultancy I'd not heard of before, but I assume that they were trying to produce a model of commuter traffic flows, probably for Luton Borough Council. (Well, that what I told LBC's parking warden when I persuaded him to let me park my car, free of charge, for 12 hours on a 2-hour-maximum meter!).
There are other surveys where we're also required to carry out 'occupancy counts' but we're more often asked to do 'turning counts'. For those, we monitor how many vehicles (of different types) turn in each direction at a junction. We might also be required to do 'queue length counts' at the junction. (Sometimes we're just asked to count how long the queues are, in each direction, every 5 minutes. At other times, we'll need to do it for every phase of the traffic lights).