The Second Wave Of Optimization For On-Demand Drivers

Ever jump in an Uber or Lyft only to find out that the driver started their day more than 100 miles away? It’s becoming more and more common as people increase their reliance on these ride sharing services. The driver picks up a fare from Los Angeles to San Diego, and decides to “work his way back” to the LA area throughout the day. With no start point/endpoint feature for drivers, they kind of have to wing it getting back home, or risk driving deadhead (empty) up the treacherous 405.

Have you ever been waiting for your lunch delivery only to see 4–5 ride sharing cars drive by? Have you ordered food and your $12 pad thai is the only order that the driver delivered on that trip? While waiting for your lunch that was the only delivery on the trip watching ride sharing cars drive by did you see a FedEx truck? … ok ok that last one was a little ridiculous. But you get the point. There is a ton of waste in the on-demand economy; and waste that could be eliminated with some careful planning and a little incumbent buy-in.

Online shopping and food delivery markets have been growing consistently through the past decade and are forecasted to grow further. (see fig. 1) With more efficient operations, online companies are getting more done with less, but if that’s the case, then why are we still using the same logistics networks put in place in the early 1900's? Loading up trucks at the beginning of the day, or running 3–4 food deliveries at a time during lunch rush?

Fig. 1 — Forecasted online US retail sales

We know exactly where the customer is, we know exactly where the drivers are, and we know exactly what traffic is going to be like. The little supercomputing, satellite-referencing, glass screens in our pockets can talk to all of the other little glass screens in the world — opening up huge possibilities for the future of last-mile delivery.

Inevitable Waste Due To Fluctuating Demand

Uber, Lyft, and others’ surge pricing is the most common example of pure economic pricing: higher prices decrease demand. But what if we look at the other end of the spectrum, at the times that a driver is underutilized? We’ll see so much waste that the only way for Uber/Lyft/Sprig/DoorDash Others to stay in business is to use part time workers.

Fig. 2 — An example of Uber’s demand curve on a week in Pittsburgh, we can assume that the curve looks something like this in most cities.

Surge pricing can smooth some of that demand, but the general up and down every day is a fundamental part of the on-demand model. This same inefficiency carries over to the food delivery and logistics businesses as well.

As we get into traditional logistics operations like FedEx, UPS, and USPS. We see less demand fluctuation, but greater delivery costs due to dedicated trips. If your Amazon Prime delivery is a toothbrush, there is a truck, and driver bringing that toothbrush to you for a flat rate, wherever you are. Some items are profitable to deliver, some are not — Amazon is willing to take a loss on these items to solidify out Prime addictions. But what about the logistics network? How can we make it more efficient? Where to do the paths of the on-demand economy and traditional logistics cross?

Most Packages Are About The Same Size (Except The Human Ones)

That toothbrush (or cell phone or coffee cup or whatever) from Amazon and that lunch from Sprig are about the same size, they are vastly different products, with vastly different delivery priorities — but they are more or less the same size as delivered to your door. I would be very willing to bet that 70% of all deliveries made to your door would fit in a box roughly 12" x 10" x 8".

I would also be willing to bet that you have a few boxes about this size floating around your house, many of them with little smiles on the side. VC Brian O’Malley references this disparate packaging trend (and the waste in the on-demand world) in a recent interview: http://techcrunch.com/2015/08/27/top-vc-brian-o-malley-on-the-next-on-demand-wave/

What If All Of These Boxes Were The Same?

The shipping container — poster child of globalization caught on so rapidly and so universally that it’s hard to imagine a world without insanely efficient ports and rail yards. Containerization increases efficiency to such an astonishing degree that I don’t understand why it’s not in play within the small-scale logistics world.

Short history of container shipping and globalization.

If my assumption of one standard deviation of B2C products/lunches shipping in a small box is correct, there is a huge opportunity to standardize delivery packaging/protocol, and thus open the world of logistics to the on-demand driver (and their downtime).

All Packages Have Priorities (Including The Human Ones)

When we’re riding in the Uber, Lyft or Sidecar, we’re the most important package. We don’t want to make any extra stops unless there is a significant discount for the hassle (hello UberPool and Lyft Line). But what if every on-demand driver on the road could carry multiple priorities of cargo at once? What if your trip from the Chinese restaurant to your house happened to line up roughly with a delivery that needed to go out? Even if the food delivery was a block away from your house, you get dropped off first and the food comes second — everyone saves (or makes more) money!

Now imagine another layer, one where anytime a driver nears an Amazon distribution center they stop by and grab a few last-mile packages. The toothbrushes and coffee cups are nowhere near as high of a priority as the food, but they do need to arrive before a given time. Same story as last time, the driver gets a fare, the destination address is near an address that requires a drop off. Boom — more money saved!

The real lesson here is that it does not matter what the cargo is, but rather what the delivery priority of the cargo is.

Enter The Mini-Container

Much like the containers used in multimodal shipping, we need a standardized container for the on-demand shipping world. A small, durable and interchangeable container that can be used just like a cardboard box is — but more than once.

With engineers from each of the delivery and ride sharing services collaborating, we could design a container that would create huge leaps in efficiency.

While the actual container itself is nowhere near as important as the software, there is still a real need for something that all services could used in a ‘one standard deviation’ model for shipping goods. Collaborations like this work REALLY well in software (thanks MIT for laying the Open Source groundwork), but can they work in the logistics industry?

What Do These Re-Usable Containers Look Like?

The good news is that there are some incredibly smart people in the world that design products for these purposes. The data can tell us the right size and the right materials for use, but there are a few big things that would be required if we’re going to containerize B2C deliveries:

1. Durable boxes made of recyclable materials.
2. Resealable openings with tamper-evident seals.
3. Stacking, nesting, and rack store-able for ease of use in-store and in-transit.
4. Cheap enough to be lost occasionally, but not cheap enough to be thrown away.
5. Individually traceable with ID numbers and huge easy-to-scan codes for the driver’s smartphone.
6. Antimicrobial.

Personally I think a tote-style plastic box with a zipper and tamper evident tie would work perfectly. (fig.3) Though there is a 99% chance that I am totally wrong (remember how many smart people work on this problem every day).

Fig. 3 — Standardized flip-top plastic boxes paired with tamper-evident zippers.

These little boxes would slide into racks in the trunks of on-demand delivery cars. Since none of the cargo is extremely fragile (great packaging) there is no need for extensive custom systems in each car, just a simple aluminum rack that the boxes slide into for transit.

The environmental impact of the huge amount of cardboard boxes is a waste without a doubt, but it is nowhere near the waste of gas/diesel from vehicles idling and making redundant trips. Simplifying this process now when the world is still using human-driven cars is the first step, bridging the cap for an easy transition into the robot-powered logistics network that we all know is coming.

The Math Behind This Isn’t Easy

Every mathematics and computer science student has worked in that hairy area known as NP-Hard problems (fig. 4), and we all know that they are nothing to scoff at as the number of variables and the number of points increase.

Fig. 4 — NP-Hard problem and solution. Dots represent stops to be made, and the red lines are the most efficient path between all of the dots on a single trip.

As the number of points (delivery stops) and and the number of variables (cargo priority, traffic, distance from home base, etc.) increase, a strictly efficient solution disappears quickly due to the limited computing power in the world:

As we can already see with these small numbers of cities, the number of paths grows extremely quickly as we add more cities. While it’s still easy to take a given path and find its length, the sheer number of possible paths makes our brute-force approach untenable. By the time we have 30 cities, the number of possible paths is about a 9 followed by 30 zeros. A computer that could check a trillion paths per second would take about 280 billion years to check every path, about 20 times the current age of the universe. Read Full Article Here.

We know that there will never be a perfect solution. But this isn’t mathematics, it’s business. We don’t need a perfect solution, just something that is better. I have total faith that there are more than a few engineers out there that could build something better.

Why Containerization Won’t Work

There are numerous reasons that everything I’ve talked about won’t work. But the biggest reason that I see is that each company wants to control it’s full fulfillment stack. Totally understandable, but with the right efficiency increases I’m sure these firms could be convinced otherwise.

A few other things:

1. Using the same containers for food and for cargo. Obviously secondary packaging is needed for food. (Imagine opening a soggy Amazon box filled with drunken noodles!)
2. Can’t find the correct balance between low-cost and durability of the containers.
3. Contractors don’t do a good job with the delivery of goods with different priorities.

It’s Already Happening

You get in an Uber and there are three phones sitting holstered on the dash — as you get in two of them get turned off. When you get out, they come back online. Regardless of the insanity of using three separate phones at once, 75% of drivers drive for more than one service at once.

Din is using the existing on-demand infrastructure to deliver their ready-to-cook meals in minutes.

Also, one of the big guys has opened a cargo delivery system. Sidecar launched Sidecar Deliveries. Using an optimized priority based system, they have taken the first HUGE step towards the standardization of last-mile cargo.

fig.5— Look at all of that beautiful efficiency.

Asking a bunch of companies to work together when they are fiercely competitive is a fools errand, but if the leaps in efficiency (and profitability) are there, then I think the market will shift just like the ocean freight market did years ago.

In Conclusion

Can this actually happen? I hope so. If you know someone who is working on solving these problems please comment and point me their way. I have no vested interest in seeing containerization happen other than watching markets evolve and mutate into things we never would have guessed years ago.

Big thanks to @Stella and @Meshlakhani for their input on this post.