Host Intro (Gene Kim)

All right. I am ridiculously excited for the next presentation, which is from Forto. It is a logistics tech company who set out to revolutionize the shipping industry. So Jacob Brank and Andy Duncan will describe the complexities of this industry, including procuring container shipping capacity and matching supply with demand in a rapidly fluctuating market, which is exactly what the Covid pandemic and the Red Sea shipping crisis contributed to.

So they will describe the development and impact of new tools they built to optimize some of the most important processes of the company and all of the skills and experiences required to make that happen. They will also describe the benefits they created, again, helping them not just win the marketplace, but ensure the smooth operations of an industry we all rely upon.

So, Jacob and Andy, I can see you. I can see your slides.

Jacob Brank

Hello. I can hear. Thank you.

Host Intro (Gene Kim)

Awesome. Thank you. Take it away.

Jacob Brank

Let's go. Thank you for the opportunity. I'm Jacob. I joined Forto, a logistics startup, five years ago. It was my first real job after university. Just moved to Berlin, didn't really know what startups are, so even though my job was very much focused on manual data entry, it was very exciting.

And so since Forto is a fast-growing startup, there were lots of opportunities for growth. So I quickly moved into more commercial roles with more responsibilities, and ended up, in the end, in product management. At the time of the events we are discussing, I was part of the logistics product owner teams, in the end responsible for the logistics product and the commercial success of the product. And I was basically bridging the traditional world of logistics with the fancy new startup and tech world.

Andy Duncan

Hi, I'm Andy. Over the past 25 years, I've worked in many different industries and locations around the world. A good deal of my time has been in consulting, where I've played both technical delivery and managerial roles.

I've worked for Equal Experts for four of the past 15 years. I'm currently Head of Engineering for Europe, where I support our consultants working for a number of different clients.

Jacob Brank

Okay. Forto is a freight forwarding company, a digital freight forwarder. So what are freight forwarding companies actually doing? In the end, we are aggregating services, because we are focused on shipping containers from Shanghai to Hamburg, from China to Europe. And there's a lot of steps in between. So we have to pick up the empty container. You have to bring the container to the port. After you put your stuff inside, you have to do export customs, import customs. You have to bring the container to the destination warehouse, and you have to bring back the empty container in the end.

So we aggregate all of these different steps, all of the sub-services and sub-providers, so our customers can book, in the end, a single service with us and have a single party executing it.

We were founded in 2016, grew quite fast, around 800 people in multiple offices in Europe and Asia. Since we are a digital freight forwarder, we want to use, or we are using, technology to do, in the end, a better job for our customers: having better internal processes to have a better customer experience, and give our customers more transparency and control into this part of their supply chain.

Working in an industry like logistics, there are a lot of challenges: lots of, I'm not kidding, paper-based communication, a very fragmented vendor landscape, a very old-school industry, and the problem of balancing supply and demand, which we want to talk about.

So let's look at this slightly anxiety-inducing graph. It's basically the Shanghai Containerized Freight Index, which shows the average price for shipping a container from Shanghai to Hamburg. It was pretty steady below a thousand US dollars for a long time. Then things started to happen. I joined the industry, not so much disruption yet, but as you know, in 2020, Corona started, and with it, the prices started to rise dramatically. We quickly reached four or five, six times the normal price.

Then the Ever Given got stuck in the Suez Canal. I'm sure lots of you remember the memes and the pictures, but for the industry it was actually a pretty bad thing because it kept pushing the prices up. And when in China the big ports started to be closed due to the zero-Covid strategy, for example the biggest port in Yantian, the biggest port in China, the prices reached basically the peak at 10x, 11x the price of before this crisis.

Then prices went down and everybody hoped for some quiet months and years. But then with the Red Sea crisis just recently, and the attacks on container ships and so on, the prices started going up again.

Why are these prices going up like this? What happened? In the end, it's the problem of balancing supply and demand, because demand can be very volatile in the industry. Think about Corona: everybody was at home starting new hobbies and started ordering stuff online. So the demand was going up. But supply is very inelastic in the industry. I mean, you cannot really quickly build a container ship. It takes up to five years. There's not many places that can do it. Also port capacity is very limited.

So if demand goes up, the supply stays the same and the prices go up, and then you have additional shocks to the supply side. For example, the port gets closed or something stuck in the biggest shipping route in the world. Things spiral out of control very quickly.

As a freight forwarder, we also have to balance demand and supply because we don't really own container ships. No freight forwarder does, because it's like a whole different nightmare. So there are a few shipping lines that own more or less all of the capacity on container ships in the world, seven to ten. So we buy capacity from them and sell it to our customers.

The other side, the demand side, is very fragmented. We have thousands of customers. Some of them ship less than five containers a year. Others are very big and ship up to a million containers a year. So balancing this incoming demand, the bookings, to the supply we have available as a freight forwarder is a core problem in the industry. And how do we do it?

Basically every booking that comes in, bicycles, furniture, has to be assigned to a very specific buying contract and a specific vessel that is departing. This is done booking by booking, hundreds of times per week, by a dedicated team in Shanghai. And it's a very operational process.

Each of these decisions can make or break the success of one of these transports, both from a financial perspective, but also from a service quality perspective, because these different buying contracts, they have different prices, different rates, but also different delivery times, different transit times and so on.

But how does this team make these decisions? How do they know what to do? In the end, that's the problem space we were in. They do it based on instructions from the trade management team. It's a team of highly professional experts that were sitting in Hamburg. They are the profit and loss owners, in the end, of the logistics product. They procure the capacity, they do the pricing, they do a lot of things. I was in this team.

In the end, these instructions can really make or break the business, and they all happen outside of the system. So the way we were coming up with this instruction was based on a lot of spreadsheets, a lot of emails, meetings, and so on. In one of the talks yesterday, we actually saw a spreadsheet that had three data imports from different sources and then a bunch of Excel code. This is exactly how it was.

To be fair, it never worked very well, but with Corona it really broke because we started making a lot of mistakes, which cost us, in the end, money and impacted the customer service. So we overutilized expensive contracts and underutilized the cheap ones. We gave the very scarce capacity to the wrong cargo, cargo that was not even ready or not super urgent, or from customers that were not so important at the moment. We booked on contracts that were already used up, so the booking got canceled by the other side, which created a lot of operational headache, and everybody was stressed all the time for more or less two years. So we needed a better way of doing this.

Andy Duncan

I'm going to share the story that played out over six months as we moved from that complex web of manually maintained spreadsheets to a set of building blocks that can be combined to support the decisions being made on a daily basis. I'll share how we started with a very simple model and iteratively added complexity as our understanding developed.

I'll use a much simplified illustration of our process. It's less to teach you the intricacies of freight forwarding, so don't worry too much about the details, and more to share how we work together to represent complex business ideas and processes in simple software models.

We found that it was important to speak to everyone. No one person knew everything about the process. Further, colleagues in Shanghai and Hamburg were involved in different aspects of the same process and all had useful insights.

We found that people had solved most problems in one way or another. So we listened carefully to match how people thought about things to how the model was evolving. We also had at least one occasion when we realized that there wasn't an existing solution, and we used the developing model to guide our thinking as we invented a new business process.

Something that struck me as I read through the initial problem statement was that the data being described fit a model I had observed many years ago when working for a client in a completely different industry. When working in energy markets in the UK, I'd seen how the anticipated demand was modeled and mapped against the different supply sources, some of which are more variable than others.

I realized that we can use the same approach for supply and demand in Forto's business if we think in terms of containers instead of gigawatts and calendar weeks instead of hours. The time series curves are simply a way of representing changing values over time. And when we model things as curves, we can add them together, subtract them from each other, et cetera.

For example, we might have two customers: a technology importer operating in a relatively stable industry, expecting a constant 40 containers a week from Shanghai to Hamburg. Another customer in the more seasonal home furnishings industry needs to get stock delivered in time for spring sales in Europe, with 20 containers a week from Yantian to Rotterdam over six weeks. We model these as a continuous flow of containers, and here you can see the sum of the two curves.

On the supply side, we have a long-term contract with Hapag-Lloyd running all year for only Shanghai to Hamburg. We then managed to get a good short-term contract with MSC for February and March that includes Yantian and Rotterdam. Note that it also includes Shanghai and Hamburg. We also model these as continuous flows of containers. And here we have separate graphs for different origin and destination ports.

When a booking comes in, we can take a slice of graph for the ports and timeframe in question and use that to decide which contract we're going to use. To begin, we use a naive approach where we simply use the topmost contract of the graph for the relevant route. So we have a series of bookings being called into the office in Shanghai and the need to make decisions as they arrive.

Here everything has gone on the short-term MSC contract, and this is going to cause all the problems Jacob mentioned earlier. Not only will we incur penalties from MSC for overutilizing the small contract, but also from Hapag-Lloyd for underutilization. We might also suffer problems if there's no space on the actual vessel, causing delays and additional cost.

The solution to this was already well understood, and there were spreadsheets with existing bookings representing consumed capacity, but they were not always up to date and often inaccurate. We accessed real-time data for bookings and represented them as new time series curves, which we then subtracted from contract volume curves. The rule changed to recommend nominating the topmost contract that has available capacity.

As we play out the scenario again, we see that the third booking is nominated to the Hapag-Lloyd contract, as there's not sufficient volume left on the MSC. However, when the Patio-Furn booking comes in, the MSC contract, which is the only one that supports Yantian-Rotterdam, is already mostly consumed and we are unable to nominate. We might have managed to avoid penalties for overbooking, but we've not managed to satisfy the needs of all our customers.

Again, there was already a solution to this problem. The MSC contract had been secured intentionally because it included the specific ports and was referred to as allocated to the customer with a specific demand forecast. This would be described in one of the spreadsheets. There was a problem, though. It was not always clear that this should not be used for other customers or for other routes.

So we modeled this information as a new time series curve. A contract could be allocated to a specific customer for a set number of containers per week. So in addition to consumed capacity, capacity allocated to other customers could be subtracted from the total contract capacity to determine available capacity when nominating.

As we play through the bookings again, you can see how the allocation for Patio-Furn prevents the larger Techno-Trader bookings from being nominated against MSC. This ensures that when the Patio-Furn booking comes in, we have capacity set aside, and this is a much more favorable result. Both our customers and carrier partners will be happier with how things have gone, but there's still a little room for improvement.

In addition to specifically allocating capacity, sometimes that procurement team in Hamburg wants to communicate a general change in priority. This could be for any reason, maybe trying to build a stronger relationship with a specific partner or preferring a lower price. While there had been attempts to communicate these strategies, it was difficult to articulate and often missed by the operations team.

So this was a case where we invented a new process. For a set of contracts that would otherwise be equivalent, the procurement team can arrange the contracts in a prioritized list. Of course, these were also represented as time series curves and combined with the capacity curves to improve the nomination recommendations.

Now as we play out the scenario with Hapag-Lloyd prioritized over MSC, the operations team in Shanghai has all the information they need to nominate all bookings appropriately.

We noticed an important shift during our work together. Balancing supply and demand had been perceived as an optimization problem that we needed to use computers to solve. But as you've just seen, the team were capable of solving the problems. We realized the problem was actually one of communication.

The tools we built allow a team sitting on one side of the world to clearly articulate their strategy so that a team on the other side of the world can confidently proceed with making decisions on bookings for thousands of customers across hundreds of ports and dozens of complex contracts.

Jacob Brank

Yes, and we rolled out the tool to most of our, all of our Asia to Europe business, and it basically delivered value for us. We increased the gross profit. During the rollout, we did a cohort analysis and comparing the cohort where the tool was used against the cohort where the tool was not used yet, we saw a 10% increase in gross profit, which is a big number in the low-margin business of freight forwarding. This comes mostly from just better utilizing the right contracts.

We also saw a 30% increase in productivity for the allocation-to-booking process, mostly driven by less rejected bookings on already fully used contracts. And as our very direct CPO phrased that, the capacity management tool removed the guesswork from the allocation and nomination processes and enabled us to make, in the end, better decisions on behalf of our customers.

And we also learned some things along the way. If you work in a very traditional industry like freight forwarding, you always think that your problems are very special to you and to your industry. So having an outside team coming in, applying knowledge from other industries, as we heard before, that have quite a similar problem and found a solution to this problem already, is super helpful, especially since other industries seem to be a little bit more advanced in their digitalization effort than freight forwarding.

And the second big learning, and maybe for this audience it's not such a big revelation, but MVPs really do work, because we turned this abstract idea of a capacity optimization tool into a very usable functionality that is released to production and that is used hundreds of times every week. So we created a feedback cycle with the actual end users that led to ongoing and multiple improvements. The tool is now still being developed and constantly improved by a dedicated team.

We've seen some shift in market realities in the last year, moving from a supply-constrained market to a more demand-constrained market. And the tool still works. We could adapt it.

There are still some problems to solve. So getting our supply and demand data in our system is still a challenge. Shipping lines are sending us the rate sheets as Excel sheets. Customers send demand forecasts as Excel sheets, emails, PDFs. I've even seen some WhatsApp voice messages. So getting this into our system, for example using AI-improved OCR, is the big topic for us still.

We now want to focus again on optimization, creating system-driven instructions for allocation based on user-generated context, but also AI-driven market forecasts. And we want to get even closer to the supply chain of our customers, tapping into the communication with their suppliers to have a better and earlier picture on upcoming demand.

And what I really like is that we can basically ask questions to the audience here. And I was thinking, okay, what's the problem that keeps me up at night? And from a product perspective, it's really that we are now building a system where humans and AI basically work next to each other on very similar tasks.

So I would be super interested to hear some examples on how other companies are going with this topic. How do UI and the UX user experiences, because this is something, I think, a challenge that a lot of other companies have, and I think it's a super interesting development.

And that's already our little talk about freight forwarding. Thanks for the interest and the opportunity.

Host Outro (Gene Kim)

Thank you, Jacob and Andy. And congratulations on the new role, Jacob. Well earned.