Million dollar gas hack.
greenwala.com
greenwala.com
If shippers provided UPS with an email address for the recipient a notice could be sent by email the day before a package is to be delivered. This notice could provide options to let UPS know to wait a day or two to attempt delivery when someone will be there or redirect the package to a commercial address (e.g. the recipient's place of employment).
I recently had a package delivered that required a signature. After the first attempt I called UPS and had it redirected to my work address. Unfortunately they already had it on the truck so I was told they would attempt delivery to my home address the next day before redirecting it.
The UPS delivery guy actually came over 4 hours later when I was definitely away from home to put up a delivery attempt notice saying I wasn't there. After I explicitly told two separate customer support folks NOT to deliver the package after I talked to them because I will not be there to pick up - and those two were separate from the initial call I made. This guy seemed to have an absurd sense of humor, because he marked the one and only notice I received that I didn't want as a second delivery attempt.
Your idea will only make sense if UPS actually did their job. They failed catastrophically for me in that incident. Since I seem to get delivery attempt notices quite often, I wonder just how many times they actually stopped by (unfortunately I usually take the car out when I'm going, and that's a different exit..).
I'm sure the system is smart enough to not have them go all the way around the block to backtrack. It just has them hit that location before advancing. I would love to see an illustration of the type of route it outputs.
Yes, I just wonder how much less efficient "the most efficient route with only right turns" is than the true optimal route. Based on my limited experience with optimal TSP problems, I bet they're fairly close.
I'm imagining a route which gives each driver a chunk of several blocks, then has them hit all locations within the chunk with a right-hand spiral inwards... But I'd love to see the actual routes.
This was a little beyond "some analyst"... even if it was originally the idea of one person, the actual plan took many, many people to implement. Solving real versions of the traveling salesman problem takes a lot of manpower. UPS hires more Industrial Engineers than anyone else.
I would first presort the stops in clockwise order around their common centroid, then optimize it with a stochastic algorithm like simulated annealing. Assign costs separately to distance travelled, right turns, and left turns... something like 1 point per kilometer, plus 0.5 points per right turn, plus 2 points per left turn.
Anyone with more experience with optimization problems, feel free to show me up :)
For example, UPS knows that turning left from Main to Elm takes 45 seconds while left from Main to 2nd takes 15 seconds. UPS knows how fast you can actually go on different streets. And, it knows how these things vary with time of day and day of week.
Hmm. UPS could probably make money selling traffic and planning information.
This is one of the things that Industrial Engineers actually do. In my opinion the pay is much too low for how hard the math is. Lots of linear/dynamic programming + statistics and all sorts of fun algorithms.