[psunavy03]

3) Amazon grad takes hold of a freight forwarder, and then realizes freight forwarding is not a tech company, nor is it just warehousing and distribution.

And then realizes his competitors rely as much on grizzled logistics veterans and their corporate knowledge about customs brokerage and how to Get Shit Done and Get Shit From Point A to Point B as they do on a cool tech stack. And then goes "shit."

Don't get me wrong. From a tech perspective, a lot of Flexport's competitors could use a size-18 boot in the ass. But tech is an enabler for them, not the core business. So there's only so much willingness to fork out cash for what senior leaders see as a supporting function. I mean, if their 40-year-old COBOL backend still works, they DGAF that it's outdated. At least until it can't keep up with the modern side or it starts falling over when all the devs retire, which is a separate story.

[tschwimmer]

This is a very nice sentiment but the truth is that the established forwarders operate in a way that is could be greatly optimized. Established forwarders (and Flexport) have teams of people that "move" shipments by applying an SOP (basically a written algorithm for advancing the shipment to signals that come in from a variety of sources (an origin factory various carriers like truckers, airlines and shipping lines, governments, etc). Paying people to do this work is expensive, and the people, while very good, go home at night, take vacations, make mistakes and are generally human. The technology that enables this process at legacy forwarders is as you say archaic, but it is archaic in _design_ as well as execution. It assumes that human beings will be sitting behind a keyboard pushing buttons and making phone calls based on what it says.

In my opinion, there is nothing about any of these processes that in aggregate be worse if they were automated, and no fundamental issue with automating them. Moving freight is about as complex as any other inter-business process in that the input is some signal (usually an email but potentially a phone call or a fax) and the output is also usually an email or a phone call. The internal logic to that process is not particularly complex either.

So there exists a massive opportunity to provide roughly the same service that existing forwarders do at a much lower cost basis. The requirement to do so is to aggressively automate operations. There's also an secondary opportunity to provide better quality of service (more transparency, more programmatic logic related to movement of goods) to customers (supply chain professionals at F500) but I think that's both harder to do and customers are less sure about how they might take advantage of that better service.

Quite frankly, I don't buy the argument that "grizzled logistics veterans" hold the secret sauce to make this work. I was told probably a dozen times that what I was doing was impossible to automate but when I dug into it, what we were trying to do was a bog-standard business process dressed up in domain-specific jargon and 'tradition' (e.g. we can't do it that way since we've always done it this way). I'm not saying that I'm some logistics savant and there's a vast amount of expertise I don't know. What I am claiming is that I've pulled back the curtain on logistics and the wizard is just a regular guy.

Source: Former Flexport employee

[sharkweek]

> I've pulled back the curtain on logistics and the wizard is just a regular guy.

I agree, but land at a different conclusion.

I dabbled in logistics a bit at one startup and the thing I found true is that every warehouse and receiving dock did things a little different and sometimes that thing was computer software and other times it was a person with sticky notes all over the walls of their tiny office by the bay door who was on 2-3 phone calls at a time screaming at/being screamed at by people moving goods around the city/state/country/world.

Logistics has a ton of room to optimize and I’m sure we’ll continue to see billion dollar startups roll out every year solving some part of it, but in the day-to-day operations of supply chain, change is so incredibly hard. The middle layer of management always seemed so risk averse to making any changes that had the slightest risk of disrupting their existing workflow because if one day they goof something up it quickly spiraled down the chain.

[monero-xmr]

In the show The Wonder Years, the dad works in freight logistics. Watching him go into his office and slowly get more insane as the day goes on is how they explain why he’s a grumpy guy.

Skip to 2 minute mark https://youtu.be/dXfFvdvi7Z0?feature=shared

[psunavy03]

All of this.

[psunavy03]

You're not technically wrong, but it's also a much more complex and ever-changing problem set that you're giving it credit for. If it was so easy, Flexport stock would be the next Amazon. What they have going for them is a tech stack that doesn't have 40 years of accumulated crap.

But your description of what goes on at the branch level smacks of someone who has a CS degree and thinks that this means they can engineer the world. Dave Snowden of Cynefin fame has some choice words regarding software engineers who don't realize that human endeavors have what he calls "messy coherence," and think they can optimize the "messy" out of it, forgetting that they're dealing with humans.

There's a lot of ways Flexport could capitalize on the silliness of some of its competitors. But its competitors are also not staffed by idiots, and if it was so easy to automate everything, it would have been automated 20 years ago.

[tschwimmer]

I sat with the Flexport logistics teams for months on end. I watched them move hundreds of shipments. Maybe they concealed all the secret sauce for when I got up to the go the bathroom but nothing I saw was particularly notable. They had rules like "make sure the ISV is filed at least 3 days before expected arrival" and "if the origin agent hasn't picked up the can after more than 2 days, call Joe at the destination warehouse and tell him that it will be late."

Designing these rules are not something that computers can handle particularly well (though Generative AI might change even that), but applying them consistently is something that computers do way better than humans (for the reasons that I mentioned in my previous comment). Yes, there will be exceptions. That's okay, we can also build in logic to handle exceptions.

The fundamental process flow we're trying to automate here is not self-driving cars. It's not cognition, perception, or reasoning. It's not even really complex decision making. The inputs (while not perfect) are actually pretty well quantized. People love to handwave and say that it's so hard and that it's naive to try and reduce it down to a computer program, but I've never had anyone compellingly explain __why_.

For what it's worth, this is not a minimization of the work done by my former colleagues and similar professionals. My colleagues were smart, hardworking, mission driven and gritty. It's just that the work that they did could be more effectively done by computers. That's all.

For what it's worth, I don't have a computer science degree. I studied one of those useless underwater basketweaving majors. That must explain the overall sense of confusion.

[psunavy03]

Then why has it not yet been done? Has God touched Flexport on the shoulder and said "We deem Ye worthy to redeem the freight forwarding industry?"

It seems to me that if it's so easy as you describe, and yet it hasn't been done, than maybe it's not so easy after all. Perhaps not impossible. Perhaps doable. But perhaps not that easy. People who work in the industry are not stupid. They don't want to do unnecessary busywork. What are they missing? And why has Flexport not taken the industry by storm if it's so easy to automate?

[mike_hearn]

Do you have experience with enterprise software projects?

For those of us who have, it's a bit of a strange assumption to assume that if something can be automated well then it would have been done already. It's a bit like asking why Blockbuster didn't set up a streaming service. A lot of companies struggle with IT projects. They avoid them if possible. They outsource if not. If they do succeed, then they try not to touch the results more than necessary.

This approach leaves a lot of automation on the table. If you're able to repeatedly pull off complex software projects well, then you can just do it over and over and end up automating far more than other companies will even true.

> why has Flexport not taken the industry by storm if it's so easy to automate

Hasn't it? From Wikipedia, it "has annual revenues of more than $3.3 billion" and "In August 2018, Flexport was named the 8th fastest growing company in the United States". Now maybe they are doing an Uber here and selling below cost services, it's hard to say because they're private and keep raising big rounds so we don't know how profitable it is. But they're clearly not wanting for business.

[tschwimmer]

Well, I didn't say it was easy. :) It's a challenging software problem and the final 20% I imagine will be exceptionally difficult to automate. I have some other suspicions about why Flexport hasn't done this but I'm too far out to say anything with authority.

As for why the the industry folks haven't done this already, that's an easier one. For the operations people themselves, you're asking them to automate themselves out of a job. The efficiency and the cost savings aren't coming from nowhere, they're coming from salary spending on man hours. And I'm not just saying this is a cynical, self-preserving sabotage - I suspect many professionals in the space don't know or don't value automation. It's not "busywork" to them, it's _the job_ and that work is the core value that they add, so why would they change it up?

Leadership at these companies are probably more incentivized to do this, but to your own earlier point, they don't have the experience in building a modern application. They don't have the talent pool and they're probably afraid of "rocking the boat" too much because they have a good business already.

As I said before, I don't think any of the folks working in this space are stupid or incompetent. But there's a good quote from Upton Sinclair that I think is relevant here: "It is difficult to get a man to understand something, when his salary depends on his not understanding it."

[jmye]

Is it me, or are you just both asking the same question from different angles, and not really disagreeing at all:

Why hasn’t anyone done it? What specific things have kept Flexport and everyone else from automating more things?

I think those are very interesting questions.

[hef19898]

Very short answer, which is also a question: What makes you think those things aren't done?

Short answer to all your questions: it is a much a people business as it is process, tools and physical goods business. Legacy systems. Global customs. Auditability. Existing operations. Global compatibility of documents.

Long answer: Your local university library probably has some shelves of books on the topic.

[uaeu321321]

1. You seriously think people with 30-40 years of experience in this area want to automate themselves out of a job?

2. It hasn't been 100% automated because 100% automation is not the goal (or reasonable). "Most things are automated" is already a huge improvement anywhere. What do you think "it hasn't been done"? Are you assuming those overnight shipments are being done entirely by humans with pen and paper?

3. Companies that need this kind of shipment are huge. Huge companies are slow for obvious reasons to anyone that has worked in any enterprise. They don't change quickly. Sometimes they don't change at all and have all market inertia in their favor.

It seems to me you're not experienced with huge operations. It's fine and usually people assume it's way easier than it looks. I'm surprised you assume it's way harder and can't be done at all. Don't bet against software, we have decades of examples showing it wins in the end.

[projectazorian]

I don’t doubt that there is a real market for Flexport’s product.

But I wonder if Flexport has achieved market saturation for the easy-to-automate work, and doesn’t yet have a compelling offering for companies with more complex logistics needs.

[moneywoes]

May I ask what are you doing now?

Do you still see room in the space? Would love to collaborate ( have an Eng background and affinity for sales)

[tschwimmer]

Huge. Gobs of opportunity. One failing of Flexport is it's trying to tackle too much at once. There's a lot of depth to each of the steps and you could build a very solid business out of automating many of them. I seriously looked at a few things in the compliance space and I still think there's a huge opportunity. Sales in the space is generally very difficult as decision makers are slow moving and conservative. But the opportunity to provide value in many places is real. I'd focus more on the supply chain guys at the big CPG companies vs the forwarders if I were to do it again.

What am I doing now? Seems like the same thing everyone else is - Working at BigCo for more money and less stress :). Not looking to change at the moment but urge folks to look into the space for sure. It's a lot of fun and very gratifying. Pure software is more boring in comparison.

[psunavy03]

Honestly, I think if someone could get really really good at one niche aspect of the problem and not do it in some obscure stack that's hard to integrate with a bog-standard early-2000s-at-best tech stack, they'd be snapped up in a heartbeat by one of the major players still trying to migrate.

Because I can name at least two times that's happened. The big guys' main concern is being able to keep a unified view of who's shipping what, where, and to whom worldwide. They don't own planes, ships or trucks, so their rep rests on stuff delivered on time to the right place and always knowing where it is.

[hef19898]

Two mistakes here in your thinking:

- The 2000s stack, honestly doesn't exist anymore. And where it does, it does so for reasons no start-up can work around. this is the field of specialist consultants.

- Doing only one thing really good is not enough anymore in that field. not even close. And the bar for good is already so much higher than seem to appreciate.

[hef19898]

How much do you know about zje logistics software ised by, e.g., DHL and the like?

Because I worked at Amazon, with DHL among others, and I can tell you that in some use cases, DHL has bettwr logistics software than Amazon does. And CargoWise is just a straight genious software suite to run everything from Customs over air, sea and land freight to warehousing. Guess who is using that, or something similar, in some form. Guessed, the majority of logistics companies.

No idea where this "incumbants bad, disruptive start-up so much better and easy" mentality comes from so. If you go into the field of logistics and supply chain with that attitude so, you will fail. And hard, and burn a ton of VC money while doing it. The only company that managed not to fail was Flexport, and even they fell short of theit lofty pitch goals.

Thaz being said, Dave Clark had, to my knowledge, not a lot to do with the "tech" part of Amazon, he always was an supply chain / logistics ops guy. And a good one at that. If anything, he treated it as a numbers driven thing, and definitely not a tech stack driven one. Part of the reason why he was good at his job at running ops.

[AdamJacobMuller]

Unsatisfied with their IBM mainframe which had a 99.99% uptime (seriously how much missed opportunity cost is that hour of downtime per year!) with a kubernetes microservice application where each of the 1000 components had an extremely impressive 99.999% uptime.

[psunavy03]

The problem is that the legacy applications running on that mainframe are tightly-coupled and thus a bitch to test. They're also hard to automate and hard to scale, because design choices made in the 80s and 90s based on people using them at the speed of manually typing on keyboards fall down when you hook up APIs to them for automated testing and for connections to modern applications.

And most of the people who know how to develop for them are about to retire. When was the last time you heard of a modern CS or IT degree program offering a COBOL elective?

And they're not easy to migrate, because you have 40+ years of business logic locked up in that legacy COBOL application, many of the developers of which are retired or flat-out dead. Did they document it perfectly? LOL! And oh, we're dealing with customs brokerage around the world and also potentially accounting info for a publicly-traded company, which means if you fuck up badly enough, people potentially go to jail.

[robertlagrant]

It's easy to integrate green screens! Just write a program that maps out all of the UI interactions on the fixed-character screen and get the computer to use it!

(For those who think this is a joke: it's not only not a joke, it's one of the better ways to roboticise mainframe interactions.)