WORLD WORLDGoogle maps has a lot of things going for it. Backed by THE defacto tech giant gives it a level of centralness that is entirely unmatched. Though apple products push you to their versions of this software, all google search will push you through their maps. Google maps in this way can be the lynch pin for many companies life or death. Having good Yelp rating might not be enough if the first taste of public opinion in the form of the google rating, end the thought before it begins.
If your hours are listed incorrectly, you can expect for customers to show up expecting service, and blaming you if you didn't take the time to correct them on what is understood to be the public repository of data that has the most up to date information about PLACES in our world.
My professional career has been entirely spent in the implementation of mathematical optimization techniques effective us in the real world. Through my time I have found that the "effective useage" is much easier said than done. You may think to yourself, "why would it be hard to optimize all parts of life, and who would want to do exactly that?" but the truth is taking the insights from optimziation models and implementing them in the real world has a sizable disconnect. If we were robots waiting to get instructions from our operators we would be in perfect pawns for the optimizations whims. Though if a task is simple enough for that, there are likely simpler ways to optimze that process than using sophisticated models running continuously.
So how can we use these models effectively, taking into account the human element that can't be, and in most cases shouldn't be optimized out of it? Look no further than the worlds best web application to see exactly how we can do
When creating a tool that uses optimization to help people make more effective decisions, time scale is one of the core dimensions of your problem. For most organizations there are three effective time scales:
This is usually mirrored in the structure of the organization:
Frequently these time scales have different requirements for resolutions as well. Most of the time it is not practical to use the same model for all three of these domains because usually there is non-linear scaling of algorithms solve times with the time dimension. Luckily the larger time scale decisions usually don't need the as high of resolution to still be useful=
I recently bought a house. In this process my wife had to get a new job because she is a teacher and needs to be close to her school. While she was interviewing for positions and reciving offers, how long her commute is played a big role in her decision. To do some comparitive analysis we used Google Maps future departure time feature.
This feature allows you to set an arrival time at some future date and get an estimated driving time for your trip. We wanted to see how long her morning and afternoon comutes would be to and from the cities she had job offers in:
Of course commute time isn't the only consideration in the job decision, but having the abilitiy to use googles optimization algorithm gave us the abilty to understand the differences between locations and the relative weight of those differences.
Because these are future trips that can change quite a bit day to day, the accuracy of the drive times isn't as important as it would be if we wanted to make sure we go somewhere in the perfect amount of time...
Before the wife and I moved, we living in Jersey City. We would frequently commute into New York City and it would often take us to trains to get to our destination. In the day of decisions, it matters exactly when the trains are arriving, and those arrival times relative to other trains.
I would map out our destination in advance to know when we needed to be at the subway, and chose which path would be the best one given my understanding of certain trains, their stations, and their general reliabity. This would also give the best ETA to let others know if we were running late or if we would need to kill time before our event
While living in Jersey City, Shelley and I furnished much of our apartement using Facebook Market Place to get good deals on used furniture. This was a great way to save money, but it meant we had to navigate the NJ highway system, something driving perfectly gridded Colorado didn't prepare us for. Google's realtime navigation is perhaps is most complete feature, helping with understanding multi lane turns with later turns taken into account. If I didn't have this navigation HUD I wouldn't have been able to safely drive all around NJ.
This technology arguably enabled the boom rideshare apps. If there wasn't such an easy to use tool to help navigate all around a city, it is hard to imagine being an Uber driver would be possible for such a wide range of drivers.
Not only do each of these tools present themselves seemlessly in the same web UI, they also allow you to plan a future trip, view the trip before you begin it/chose the specific route, and start you in the real-time navigation mode. Each of these decisions domains seemlessly transitioning into one another to take the strategy, to the plan, to the execution all within the same framework
It is important to see that Google maps has so much more inherent value than just the navigation. While as an optimization expert, I tend to focus on those aspects of the tool, the map, the data it presents and how you navigate it are arguably much more valuable than the intelligent navigation that goes along with it. With all of the tools, seeing the state of the system is a prerequiste to advanced decision making.
When I think about making useful optimization tools for real people in the real world I try to keep these domains of decisions in mind, and the shortcomings of these tools. Each application has built in user freedom, Allowances for me to make higher level decisions outside of the tool. They allow me to depart from the plan without completely crumbling, and they give me the ability to understand the effects of my naviagation decisions in a way that would be impossible without them