The emissions conversation usually starts with vehicle replacement, which is the slowest and most expensive lever available. Four things move fuel burn with the fleet you already own, and all four cut cost at the same time.
Why start with idling?
Because idling is fuel burned to produce nothing, and it is concentrated. Pull idle time by vehicle and by shift and the pattern is rarely spread evenly: it is a handful of vehicles, often a depot habit, sometimes a legitimate reason like a cold cab or a PTO run that nobody had documented. All three are actionable, and none require new equipment.
How much does driving style matter?
Enough that it competes with vehicle choice. Harsh acceleration and late braking cost fuel directly, and cost more indirectly through tires and brakes. This is the same coaching loop that drives safety improvement, which is convenient: one program, two results. The mechanics are in how you actually change driver behavior with telematics.
Where does routing fit?
The saving is in the miles not driven: the trip that was assigned to a vehicle 20 miles away because nobody could see that a closer one was free, and the second trip caused by a missing tool or a failed first visit. Both are visibility problems rather than optimization problems, which is why they are usually cheaper to fix than a routing project.
Does maintenance really affect emissions?
Yes, quietly. Underinflated tires, a fouled filter and an engine running on a fault code all burn more fuel than the same vehicle serviced on time. Servicing on engine hours read from the vehicle rather than on a calendar date keeps the interval where the wear is.
How should you measure it?
From fuel or energy consumed, not from distance travelled. A distance-based model cannot see idling, and idling is precisely the fuel you are trying to remove, so a distance model will report progress your fuel card does not agree with. Measure the input, and the reduction is real.