Episode 1: Nidhi Gupta

Chapter 1

Intro

We are welcoming in Nidhi Gupta, she is the Co-founder and CEO of Portcast which provides artificial intelligence and machine learning for the logistics and maritime industry. And this podcast is actually our first episode of Maritime Means which is a podcast by Spire dedicated to building a community of innovators, so, Nidhi welcome into the very first episode of "Maritime Means...", we're happy to have you.

Yeah, I'm really glad to be here. Thank you for having me.

Chapter 2

Startup scene

With the government promoting and investing in technology startups, Nidhi discovered Singapore as the greatest location for her startup company because it is home to many experienced and skilled individuals.

You're primarily focusing on the forecasting and the data side of things. But before I sort of get into the whole data discussion, which is primarily what we're gonna be talking about today - you're located in Singapore. And, as I was doing a little bit of research for this interview, as someone who's based in the States, as a newbie to global shipping, I'm sort of fascinated by the bigger ports in the world; and especially when it comes to some of some of the inefficiencies on the US side of things, and how efficient other ports are all across the world, including the port of Singapore. What does the startup scene look like in Singapore?

Yeah, That's a great question. Actually, Singapore has been really fascinating from a startup perspective. In fact, I was reading somewhere that there are about 15 unicorns in Singapore now, and a lot of investment from venture capital firms, especially from seed to series B or C, I would say has been happening here. It's kind of the hub for Southeast Asia anyway, and that kind of allows a lot of startups to be based here.

On top of that, I think the government has been just really hands on, just really promoting tech, both bringing in international VCs, but also promoting deep tech startups and investing themselves. And I think ultimately a startup can grow if they find the right talent, and Singapore, with its multinational talent and people coming from a really experienced skillset across the world to be located here, I think that kind of offers it a real advantage in terms of being a startup hub.

Plus, for Maritime and logistics, for a company like ours, it offers both the corporate side, as well as the being the largest port in the world, it's just a great opportunity to start the company here.

Chapter 3

Importance of Asia Pacific

Asia serves as the world's factory. A lot of what we see around us - in fact, 90% of what we see around us - moves on ships, much from Asia to European or US trade lanes. However, intra-Asian trade is equally important.

Now from from a global shipping perspective, the Asia pacific trade lanes are just so vital to global shipping in and of itself. But for folks who may not know the important aspects of Asia-Pacific, can you give us a little rundown of how important this region is to global shipping?

You know, Asia is kind of the factory of the world. A lot of the things that we see around us, in fact 90% of the cargo or things that we see around us, travels on ships. A lot of it goes from Asia to European trade lanes or US trade lanes, but intra-Asia is also a significant part of this.

What's also interesting about Asia is that, if you think about a ship - which is kind of like a bus, it goes through many more ports before it reaches the final destination where the cargo has to be dropped off - in Asia, you would see a lot more transshipments, you would see a lot more disruptions for typhoons, a lot of inefficiencies, and that makes it a harder place to really figure out what's happening to the supply chain, but this is such a core aspect of trade lanes and import-export. So yeah, so this is where the majority of the cargo movements kind of happened from.

Chapter 4

Trade lanes

Based on supply and demand, trade lanes were formed from countries’ import and export patterns. Routes and ports would be established by shipping businesses to serve countries. Several factors, such as economics, weather, and other geopolitical issues, must support trade channels.

Now, I heard you mentioned trade lanes a couple of times, and that was also mentioned in some of the research that I was going through, and as a newbie to this industry, I was wondering - how are trade lanes established? Is it geopolitical, is it all of the above, give us a little bit of insight on that?

Yeah, I would say it's all of the above, I think at a very basic force principles kind of thing. It's all about import and export patterns within, cross-countries, and it's based on the supply and demand. So, based on that, shipping companies would set up the vessel routings to serve countries connecting the port of loading and discharge, and each vessel would kind of pass through multiple ports to ultimately serve those lanes. There would be some lanes which are higher traffic lanes, which would be, let's say, China to US, or China to Rotterdam. So those would be higher traffic lanes.

Again, based on where the manufacturing hubs are for most multinational companies and where the demand and the consumption really happens. And that also drives then prices. So the higher traffic lanes are where most carriers or most shipping companies will run more ships and would likely have lower rates to get more market share.

And then there could be seasonal traffic and disruptions, which can kind of change how new trade lanes emerge, new routings emerge, like we've seen in the last couple of years.

Well so, new trade lanes, this isn't something that is set in stone. You can establish a new trade lane pretty easily, or I imagine there's probably a lot of factors involved in that?

Yeah. Yeah. Absolutely. It would be a lot factors - ultimately it's a large, large vessel running across that particular trade lane. So it needs to be backed by the economics around it. There needs to be significant demand, the weather like you said has to play a role as well. Can the ship actually, you know, take that path? There could also be other geopolitical things like piracy, and unsafe areas, or compliance - not going to, let's say, sanctioned ports, etcetera.

So there are multiple aspects to take into account while a new trade lane is being connected, so it's easier to see new ships on routes that are getting importance rather than a completely new trailing by itself.

Are there different trade lanes for the bigger versus the smaller ships? And I think that's the biggest debate here over in the US, there's a lot of ports that don't have the infrastructure in order to handle some of these bigger cargo ships, so are there different trade lanes for different sizes of ship?

Yeah, absolutely. So a lot of the intercontinental trade, again, let's say Asia to the US, those would have larger ships, which would be the mother vessels that would be traveling between those trades. And then the smaller ports, the connecting ports, would have more freedom vessels or smaller vessels. So essentially like a hub and spoke, where you connect different ports to sort of more important ports. And the mother vessel basically starts from there. So that's what kind of creates transshipments as well.

Chapter 5

Data in market volatility

Because of competitive differentiation, Nidhi believes that the shipping industry has been somewhat opaque in sharing data. However, she feels that more chances for great value arise from data-sharing and availability.

I love that - it's so fascinating to me on how these things get determined and how they change over time. So thank you for breaking that down for me. There is one of your quotes that really stood out to me that I absolutely loved - I feel like it should be in marketing material everywhere; you said "helping shipping companies make money, not by accident, but by analytics". I thought that was such a boss quote. So let's go ahead and dive into sort of the the data side of things, which is really what Portcast specializes in, and what you specialize in.

And during an interview for entrepreneurs, entrepreneur first investor day, it was four years ago, you were preaching the importance of using data to make educational decisions, to turn market volatility into a competitive advantage. It still feels like we can make that pitch today. A lot of companies now - why do you think that they have hesitated in that four years in not making data a priority?

Yeah, that's a great question. The shipping industry has largely been a bit opaque about sharing data, and I think that that is where the opportunity has arisen as well. That's how profit was generated. Also, the largest companies, there are fewer of the largest shipping companies and they've kind of kept the data really close, it's seen as a a competitive differentiation.

However, I think what's been happening over the last few years, or even the last decade, is that there's a lot more public data now available. So companies like Spire, for example, sharing satellite data, or even going down to nano satellite data; or better information; or getting data directly from not just shipping companies, but also freight forwarders and from terminals and ports. I think the availability of data has been a lot more. I would say it's even been a 10x more opportunity in terms of data availability, and that's kind of transformed the companies which were hesitant about sharing this data.

And they've started to see that it can be a competitive advantage as long as you share it in a secure manner, and you make sure that the confidentiality of the data is not lost, and then you can create immense value from this data. So I think that's emerged and more and more companies are now accepting it in the space.

Do you find that companies are probably hesitant to share their own data, or can they be anonymous when they're sharing their data? How do you alleviate some of those concerns?

I think that the key is the business case about what you do with that data. Like - why do you share that? What is the value you would generate if you would share that? That needs to be crystal clear.

Secondly, a lot of companies do feel that the internal data is theirs, they have it, they own it and they can bring smart people to create meaning out of it. What they're really missing on is external datasets. So, the last two years have shown us that solely relying on internal data is insufficient. You're basing decisions off historical patterns which may not stand true to the test of time, especially in case of disruptions. So, what really matters is combining that internal data, which is really absolutely gold for all these companies, with external data and that creates value.

And when linked to a specific use case and business case, I think that's when meaning can be generated. So I think that's really allowed companies to be a little bit more open about it.

Chapter 6

Data driven

The most important aspect of using data efficiently is determining where and how the data will be used and the success rate of using the data acquired in your aim. After reviewing the collected data, experiment in a controlled environment with the appropriate group of organizations or people to establish meaning for your business case.

Now, when you're approaching these companies and they haven't had a data plan - a forecasting plan - previously, and they want to get started, what are some of the ways that they can correct that in action and where are the first places that they should start when they want to become data-driven?

The first thing, like I said, is really about being sure about what is the end objective - what do we want to get out, and what is the business and the use case that we're looking to drive: the problem statement. So I think that's the first thing - before really diving into any kind of datasets, internal or external, the right stakeholders within the company need to come to the table and really define what's the end objective.

Second, I think - related to the objective - the success criteria of, let's say a proof of concept, needs to be made very clear. If we run a proof of concept with such and such kind of data, what is it that we want to see, which would say "okay, this is great, let's go ahead and let's really roll it out big time, which will drive value for us".

And once that's done, then, in a controlled environment and kind of a sandbox environment, there can be proof of concepts, or rather I would call it "proof of value", where one can share that internal data, it can be anonymized as well for that proof of value, and then, partnering with the right set of companies or talent who can really create meaning from it related to the business case. And a lot of companies actually try to do things mostly internally, but I think depending on the business case, there's value in either partnering with an external company which specializes in that business case, or really trying to do it internally.

So I think having that objective clear, defining what the success criteria is, and then a controlled proof of value.

Now, when you say "defining the success criteria", I imagine most of these companies, they really just want to improve efficiency and they want to save money. Are there any other sort of success metrics that they're aiming for? Outside of saving money and efficiency - which are big ones - but I was curious if there are any others?

Yeah, that's right. I think ultimately, from a business perspective, it is really top line and bottom line, so it is really about efficiency and more growth; however, it can be broken down into objectives or success criteria which point towards that. So, what's the driver of growth?

So, let's say in our case of Portcast, what we do is we predict if cargo is going to arrive on time; so, the success metrics are how good is that prediction, and then translating that into if that prediction is good enough, and, early on, if it's good enough, what does that translate into in terms of efficiency and top line?

So, yes, the goal eventually is top line and efficiency, but, for the particular dataset that's being used for the particular use case, there are success criteria which are kind of intermediate success criteria which can be tested out as well.

And so, on the Portcast side of things, when you are collecting this data, is it proprietary data, is it your own data, is it the customer's data, is it  a combination of all of those things? How are you getting these sources?

Yeah, it is actually a combination of this, a real amalgamation. Honestly when we started the company, from my perspective, I thought Machine Learning, the predictive piece, is the core of the technology. But I've come to believe that the data ingestion, the very data sources, the advanced datasets we get, that is an equal component to Machine Learning. So where we get this data from is, for example, connecting directly to carriers and freight forwarders and getting data from them. We're getting data directly from ports and terminals. We are also getting external data from companies like Spire, which is nano satellite data about where any ship is, the latitude, longitude it is at, how it's moving, the speed, direction... We also get data related to weather, down to granular information at any point in the ocean about wind speed, wave height, but also data from Met departments around the world about specific cyclones or typhoons, the intensity they're traveling at...

And then, you know, any external disruptions that are unforeseen, like events, Covid closures, specific holidays, etc. So there's a ton of data that we're basically ingesting, and I think the multiplicity of these data sources has been a key, because no single source can really close the gap in data.

And then we've kind of built in the quality and validation loops in an automated manner, but there's also human in the loop for new patterns that emerge in this dataset. So also, creating that meaning from this data, even just making all this data speak the same language, so that it becomes machine readable, I think that's been a greater challenge from an engineering perspective.

Chapter 7

Artificial intelligence in shipping logistics

Developers and data scientists are fascinated by data and how to extract patterns from it. Portcast applies streaming data to a problem case by constructing various data pipelines and processing the data such that it is machine-readable. Following that, models and predictive algorithms are created.

I'm endlessly fascinated by the Artificial Intelligence angle, and how that even is created. What does the day in the life of a Portcast employee or developer look like when you are creating an Artificial Intelligence Machine Learning? What does that look like whenever you're starting up a Machine Learning program?

That is an interesting question. I think, developers and data scientists, they are fascinated by using the latest technology stack by looking at data which is what we call "streaming data", coming at thousands of new data points generated every second. And then, how do you kind of create patterns from it?

So a day in the life of a technical person at Portcast would kind of be about looking at the use case within our larger problem statement, in terms of data science or technology, and looking at the kind of data, ingesting that data, creating various data pipelines, transforming that data so that it becomes machine-readable. And then developing models, predictable goal items about, you know - is a particular typhoon going to have an impact on movement of a large 10,000 To vessel or not? And if so how much quantifiable impact, down to the minutes, is it going to have? Not just tomorrow, but a week ahead, two weeks ahead...

So, I think, every other day there is a new disruption, it could be a political conflict, it could be a typhoon, it could be related to Covid... and that's the beauty of what the data scientists are able to do. They're able to really quantify what the impact of these disruptions is, on a single vessel and down to a single container. And that makes life of a manufacturer much easier to really receive that information and act on it.

I imagine that a lot of these data scientists are probably - if they make like a really good prediction that saves a company a lot of money, they're probably high fiving in the office. I imagine it's almost like a sports victory at that point for these guys and women.

Yeah, I think that the biggest reward is really when the customers come back and they say that they've really been able to transform their supply chain, they've been able to change their planning, their productivity has improved because of the work we're doing. So I think that direct impact is the biggest success, the biggest reward.

When you're talking about your customers, what does a Portcast customer look like? Are you out there, seeking them out? Are they finding you? Because I imagine, becoming more efficient and saving money, you probably have a waiting list outside of your door for folks who want to work with you. But what does that typical Portcast customer look like?

So we've been largely focusing on large enterprises, which are freight forwarders or shippers. We also work with some large tech providers in this space. And the kind of companies we focus on are exactly the ones who are trying to solve the problem about ship visibility and reducing the manual hours that they spend looking at cargo movement; or, if their customer service teams are complaining that the cargo is not arriving on time and they're losing sales because of that; or they have additional supply chain costs because of buffer inventory or port fees or waiting trucks. So that's the kind of customers we've been working with, in specific sectors like F&B, consumer goods, automotive, pharmaceuticals... And that's been something where we have some of the largest customers in this space.

How we've been getting these customers - actually, when we started the company, we were really knocking on customers' doors and trying to kind of prove our value proposition. But in the last couple of years, the tide kind of turned and we've had a lot of customers inbound, customers who felt this problem and want to make a difference, and therefore this has just resonated a lot more. So it's a good problem to have. We have been working with some of the great brands and names in this space, so it's been a top-of-mind problem in this space.

Chapter 8

Portcast

Portcast is mainly self-contained. As the name implies, Nidhi integrates into the systems of customers through APIs or a web interface. Customers may easily integrate their transport management or ERP system into their web interface, and it accepts shipments and containers that are traveling globally.

So when you get a new customer through the door, what does that onboarding process look like? Is it simply just plugging via an API into their different technology stacks? Is it having interviews, maybe with some of their customer service managers? What does that onboarding process look like?

I would say it's largely plug-and-play. We can plug into the systems of customers either through APIs, or they could use our web interface and really get started, and it could be as little as 1 to 2 weeks for even a large enterprise. We could even do more custom integrations into, let's say, transport management systems, or systems like SAP.

We could also allow a sort of proof of concept for the first month in this process just to see the live data and tangibly feel what the value is going to be like. So it's fairly easy to get started, it is a standard process, it's been tested and validated with large enterprises, so it's ready to go live for a customer as soon as the business case is confirmed.

How does Portcast fit into their work day? Is it a tab open on their screen? Is it regular calls with you? What does that process look like?

If they have access to the web interface, it kind of integrates into their transport management on our system and it accepts the shipments and the containers that they have, traveling globally across the world, and it constantly shares exception alerts or information on what's going to happen to those - what's running late, which are the shipments that they have to act on and change the planning, either from a tracking perspective or customer's perspective. So it's really notifications, alerts, directly into the email, or it could be shareable links and allowing collaboration between different functions within the organization.

And then analytics, to understand what the performance has been across the supply chain and how they can create better cost savings, or productivity, or work with carriers better going forward. So that's how it kind of plugs into their workflow, and in terms of the planners and the logistics professionals, it's a daily involvement and engagement on how they use this data.

Chapter 9

Monitoring disruptions

Interruptions such as traffic congestion, political crises, or the closure of specific ports can also be recorded, and that data can be utilized to inform clients of certain disruptions so that a solution can be easily provided ahead of time.

Did any of this process change because of - I mean obviously Covid has sort of thrown a wrench in everybody's supply chain planning. How did that affect your overall data collection and planning? Did it just force you to start from the ground up, or did you have to add in another element as far as the data analytics - how did Covid affect your planning and your forecasting?

Yeah, I have to say that the kind of data we were collecting was already able to take into account disruptions like Covid, or even disruptions that we've seen beyond that, like the Suez Canal congestion, or the political conflict situation, closure of certain ports, etc. So, interestingly, the data that we were collecting was already giving us real-time information about any such disruption.

So what we did, for example, is when the Suez canal congestion happened, we were able to share live updates with our customers and also in general with anyone visiting our website in terms of what ships are impacted, what containers are impacted, how much they would be delayed. And that allowed them to have better visibility, even if they were not getting that from their logistics partner or carrier directly.

Similarly, in case of the Ukraine/Russia conflict, we were able to say what kind of vessels are likely to be impacted because of that at the very start of the conflict. And that's just the data that has allowed us to do that, so we have not had to add new data sets, although we're constantly finding multiplicity of the same data sets, and that's where the human in the loop comes in in our data validation processes - there's something missing? Are there patterns that we did not recognize and that we can add from new data sets?

So, one of the more recent segments of data that we've been adding is getting data directly from terminals; knowing what's in this black box of the port and terminal, once the cargo arrives there, and getting more visibility in that aspect, both at the origin and the destination. And I think that makes it a really interesting part of the entire ocean supply chain.

Fascinating to really dive into the fact that you already had a lot of those datasets, and a lot of those customers of yours were already prepared when these different crises hit. I think it's fascinating - and you were talking about, with the ports, and with the port of Singapore, that - one of the crazier stats that I saw is, every two minutes a ship arrives or leaves Singapore. So with that much data that you're processing, how do you prioritize what to look for? Do you have these data points that have been set in stone for a while and you wake up in the morning and you already know what to look for, or do you let this sort of data tell the story for you?

You know, you mentioned Singapore. We do this globally and at scale, so our technology works - let's say there are about 18 million containers anytime across the oceans, so 18 million containers across, let's say 6000 vessels, about 2500 ports and terminals - anywhere in the world. That's kind of the magnitude of data that we're ingesting, multiplied many times over. So the technology, the way it works, is Machine Learning, so it's not rule-based, it's not static in nature. The Machine Learning allows the system to find patterns, and there are validations that are then built in; these patterns are dynamic in nature, so trends that we see, let's say last month, would be very different from trends that we were seeing two months back. And the machine learns over time and changes and adapts directly.

So there is very little rule-based modeling, the rules of validations come in only to make sure that the data quality is in sync. We own data quality, we want to be the most trustworthy source for ocean supply chain data and that's where the validation is kind of coming. So everything is really dynamic in nature.

Are there different data points, port to port? Or is it essentially kind of the same data points that you're looking for at any ports all across the globe?

Yes, I mean, each port has different patterns - it could be about the efficiency of the port in terms of the waiting times, and the traffic at the ports. These are things that define congestion levels, but then there are also seasonalities; sometimes if we see high traffic or high waiting times, it's just a function of seasonality, and it's not really congestion as sometimes news tends to call it.

On top of it, there are different queuing patterns - sometimes the first ship coming in doesn't necessarily get the first right to berth at the port and, same, the last ship coming in could actually be the first ship to get that berth spot. So the queuing patterns vary with every port. So, yeah - and then there could be seasonal events or unforeseen events. For example, Indonesia around Ramadan: the crew could take longer breaks, and therefore that could be a seasonal pattern; if there is a port strike that could cause disruption as well.

So, these are various datasets that have to be taken into account looking at the port stay time and what's happening at the port once the ship is arriving there.

Chapter 10

Quantify piracy and global shipping

With AI linked to logistics, Nidhi can observe patterns of normal routes the ship would take around that specific location, giving them an indicator when looking at data and combining it with news that this is happening due to piracy.

That's fascinating. And so, you mentioned earlier - I wanted to piggy back off of your answer - because you mentioned earlier about piracy being a data point, and I'm kind of just fascinated by that. I've said fascination, you know a lot during this interview, but it justifies it. How do you quantify piracy on global shipping?

We don't quantify piracy, but - okay, so how it would work is, let's say, if you take the oceans of Somalia, what we would do is we would see patterns of what are the typical routes that the ship would take around that particular area, and it would give us an indication, when looking at data and combining it with news, etc., that it is because of piracy that this is happening. And the more patterns that the system sees, the better it kind of understands how to quantify if a particular route is changing for a particular vessel.

How do you then take that in order to make it into an actionable plan for your customers? Is it notices that you're sending out? Is it shipping forecasting plans? I think that you had mentioned that the most accurate data to create a vessel forecast is one week ahead of arrival. So what are you looking for data-wise to make those forecasts?

Like I said, we're combining various datasets, so we're looking at patterns of a vessel, the metadata of the vessel - what kind of ship is it, the size of the ship, the tonnage, the speed, the typical distance and route that it should take, the ports that it's traveling at, the schedule it has, what's happening at those ports, is there congestion, what's the waiting time like, what are all the other ships doing around that region? And then any other disruptions, it could be about blank sailings or skipped boats that the carrier is making changes on, it could be about disruptions like typhoons... So those are all the different patterns and datasets we're taking into account. Yes, the predictions become more and more accurate as the vessel or the container arrives closer to the ports, so, one week before the arrival it would be absolutely more accurate.

However, what we try to do is, the predictions have to consistently be better than the baseline data that our customers are receiving from the shipping carriers at any given point of time, even 2-4 weeks ahead, we're able to show that the predictions are much better than simply relying on carrier information, which may not be real-time updated. But a lot of our customers really care about, let's say two weeks before, getting the best information so that they can tell their sales teams and their production teams, and freeze plans related to production and also customer orders. So that kind of 1 to 2 week window is what we really try to improve as much as possible.

With all of the data and the information that you've accepted, or that you're creating and that you're analyzing on a daily basis - it's sort of been catapulted, I think into the forefront of a lot of different companies, especially in global shipping, that they need to start making their data actionable.

Now, and you also said - I think it was interesting that you said that Covid was an inflection point for logistics and tech adoption; now, separate from what you're currently doing, what do you think tech-wise should be adopted more within this industry? Is it more technology? Is it better decision making? Is it using technology to drive decision-making? What do you think is the next step as far as innovation for global shipping?

The biggest thing, and the biggest theme in recent times, and I would say in the next couple of years, is going to be about visibility and resilience in shipping. And the next thing beyond that, I would think, is sustainability. Shipping is a large contributor in general to emissions, and with manufacturers becoming more and more responsible, looking at scope three emissions, and sort of pressures in from all regulators for shipping companies as well, I think sustainability is going to be a really big theme and agenda.

And then, the other thing I think is about automating workflows. The next generations who are entering the space of metadata and logistics, they are used to an Uber and Amazon, and they don't want to deal with archaic systems and Excel spreadsheets when they come to work, and we need to make the industry more attractive to younger talent. And we also need to ensure that we're getting all of our goods on time and cheaper, faster, and that will only happen if workflows are automated, we're doing things not really manually or based on historical patterns anymore.

So I think sustainability, resilience continuing to be a theme, but also workflows being better and automated.

Well, I feel like that is a perfect spot to end this conversation because you really hit the nail on the head that it's not just about saving money and increasing efficiency, but if you're saving money and increasing efficiency then it has those downstream effects of helping all of us with the things that we want, faster and cheaper. Now, Nidhi, it's been an awesome conversation with you; where can folks follow more of your work, check out Portcast, all that good stuff?

Thank you. They can reach me personally on LinkedIn or our website, which is portcast.io. I would be more than happy to answer any questions or see if you can help any company get better and more efficient at shipping.

Perfect, thank you so much for being the first guest on "Maritime Means..." by Spire, and I think that this is a great way to kick off this new series. So thank you so much again for your time, and it was a pleasure talking to you and learning more about all the solutions that you're providing.

Thanks a lot, it was great to be here, thanks for the questions.

Go to https://portcast.io/ to know more about digitization and innovation in the supply chain and logistics industry. You can also follow Nidhi on her socials:

Twitter | LinkedIn