Author here! This was the first post of our new Flexport Engineering blog (we'll be adding more content soon). Happy to answer any questions about the article or how/why we're using kahn's algorithm.
Thanks for the article, this would have been helpful to me some time ago. I found out the hard way that data quality in logistics is generally quite poor.
I think the section talking about that was phrased a little weirdly. It was meant to express that we decided to assume transitivity unless we could find a counterexample, and that we haven't yet found one. I've updated the wording a bit to have it be clearer now!
Hi, I would like to share an idea i had a while back with you flexport folks.
I work in an IoT company that focus on the hardware side, while I am the one responsable for having the software running smooth.
Chatting about the new hardware piece that were coming my first though was the following.
Have a very small piece of hardware, transmitter, to ship along whatever you are shipping. The transmitter will regularly (even every 5 minutes) transmit its position so that would be possible to track the geographical coordinates of every lot of product that is shipped.
The difficult part is that the transmitter need a receiver (with Internet connectivity) in close range ~10km, the solution would be to install those receiver in the ship or in the truck or in the train.
The hardware itself is something that we are using every day and it is not a big issue, the only real problem that I see is the installation of those receivers.
From what I've seen of this industry, there's two major hurdles to getting this working: One, getting the shipping companies to provide any kind of network infrastructure would be like pulling teeth, and two, a lot of the companies doing this shipping are running on margins so tight that they care about things like saving 25 cents on a box. Their idea of cheap is completely different from yours.
I'm interested in solving a slightly different issue: could you build a cheap device that records accelerometer data while an item is in transit, to create a record of which way a box was oriented, and how hard it was dropped? Maybe provide a USB interface to download the data on arrival? Something like that would be interesting to businesses that ship fragile, high-value items and want evidence to back up the claim that items were damaged in transit. (I'm thinking of things like television sets.)
The connectivity issues is clearly a major issue, the idea would be to actually pay the carrier to install the receivers.
Honestly I don't know the industry and this is the reason why I am here asking, so please correct me if I am wrong.
I believe that shipping companies are the least interested in this product.
The one interested are definitely companies that will receive the shipped good.
Suppose Adidas, a company that makes clothes in southeast Asia.
I believe that they are extremely interested in knowing where their pallets of Egyptian cotton is so that they can have an idea of when start the production.
The whole idea get even more interesting when you add more intermediate.
The cotton is about to get to "denim ltd" so they will start the production in two days and they will ship my denim in a week from now, great I have an extra of slack for the maintainance.
There exists already mechanical drop and tilt sensors, such as [1], which are commonly used for this purpose. A search for "shipment data logger" results in some electronic devices which do what you propose, though most of them are more expensive and seem intended for reuse e.g. when shipping machinery to a location for installation.
Traditionally there are a lot of issues (battery vs. broadcast power vs. non-transmissive material (like fluids)), and it doesn't replace the need for old fashioned inventory management. At best it supplements it, and in a margins game the advantage just doesn't materialize into monetary savings.
You do see similar technologies in some situations (RFID), but solving the general opaque shipping problem is surprisingly tricky.
Traxens [0] does that. They have agreements with two shopping lines to install transmitters on their ships which then forward the data of the monitoring devices attached to the containers
This exists where the economics make the most sense. Some shipping containers have built in tracking and satellite links. Usually the ones with expensive or refrigerated stuff.
It sounds mainly like a pretty basic topological sort applied in an clever way. You probably could find pseudocode for the basic algorithm on Wikipedia.
As far as I know we don't intend to release anything. But if you have a similar timeline ordering problem, the post should contain enough detail for you to apply the ideas.
