Street-Level Routing: Myths and Fallacies
Author: Israel Beniaminy, Senior VP Product Strategy
Planes travel in the air and aren’t bothered by streets, rivers, or most other land features. Cars travel in streets and are thus constrained by the street map. Everybody knows this. However, in the domain of field service planning this obvious fact is often ignored. Read on to explore this strange impulse to overlook the simple fact that cars don’t fly.
Imagine that you need run to a few errands in an area that you aren’t familiar with. As you enter your car and start thinking whether it makes sense to go to the travel agent (A) first, then to the bank (B) and lastly to buy some candy (C), or do it in some other order, you use your smartphone to look up these three addresses on Google Maps.
However, you discover that instead of a map, your smartphone shows a blank screen with just three dots – A, B and C – marked on it. A small note explains that this change was done because you don’t really need to see the streets when deciding whether it’s better to go to A, then B and finally C, or some other route such as B to C to A. Only after you decide to start with, say, point A (the travel agent), the software will tell you what streets to take to get there. Would you be happy with this new, neat and simple user interface that doesn’t overload you with the clutter of street map details? Probably not.
Obviously Google doesn’t do anything similar to this, and neither does any other mapping service that I know of. Why would they? It’s easy to see that sequences such as “A to B to C”, which look efficient when the points are shown on a blank “map” (without any streets marked) will often be inefficient and waste a lot of time when you actually try to follow that sequence in the real world.
(Learn More: download – Street-Level Routing: Where the Rubber Meets the Road)
If you think this is obvious you’re right, but, in spite of that, many software services for planning such routes skip this obvious step of using the street map to select a good route. When asked why, the software’s vendor might offer one or more of the following explanations, which are no more than convenient myths:
Myth 1: After the route – A to B to C – is selected without referring to a map, we can refine it by using the street map to guide us through each leg, e.g. the leg from A to B.
WRONG: If “A to B to C” is a bad sequence, then there’s no sense in refining each leg separately as long as we don’t change the inefficient sequence of legs. Furthermore, such software usually plans routes for many mobile workers at the same time. If, for example, B is on one side of a river and A and C are in the other side, it makes more sense to reassign B to another worker, not just calculate the best way to find a bridge that lets the original worker get from A to B.
Myth 2: Travel time estimates are just estimates and they are never completely accurate even when using the street map, so why bother with it?
WRONG: Accuracy isn’t an all-or-nothing affair. In many cases, going for higher-accuracy estimates makes a huge difference in the outcome, even when considering that these estimates are still not 100 percent guaranteed. In running our family errands we think ahead about the roads we need to travel, and we would find it hard to accept the advice that we shouldn’t bother since we can’t really predict what the minute-to-minute road situation will be. Why should we accept this advice in running a mobile service operation?
Myth 3: When ignoring the street map, the travel time estimates sometimes over-estimate and sometimes under-estimate, so it all evens out.
WRONG: It may well be that there would be an averaging-out across the whole workforce. For each mobile worker whose total travel time was underestimated by two hours, there might be another worker whose total travel time was overestimated by two hours. This does not solve anything. The first worker would have to work two hours of overtime, while the second worker would spend two hours doing nothing but collecting pay.
Myth 4: Let the workers drive enough routes enough times, and the software will automatically learn how long it takes to get from every point to every other point, so it will learn the impact of the street map on travel time.
WRONG: First, why would you want to ignore easily available information – the street map – and force yourself to go through expensive trial-and-error until the learning process improves the travel time calculation accuracy? Second, what happens when a new road or bridge is opened? This would change the travel time across large parts of the map, so learning and trial-and-error must begin again. Third, such learning almost never really looks at all the possible addresses, since it is just about impossible to wait until someone drives from every address to every other address so that we can learn the travel time.
Instead, the learning process clumps many addresses into a much smaller number of groups of addresses and then learns how much time is required when driving from any point in one group to any other point in another group. Clearly this introduces errors. Some of the errors are small yet still harmful to the feasibility of many routes, while other errors will be wildly off.
Myth 5: Street level routing is just too complex and time-consuming to do when planning many routes with many legs in each route.
WRONG: It can be done effectively and built into high-quality, high-speed route optimization.
Learn More: download the business paper – Street-Level Routing: Where the Rubber Meets the Road.