jcpjr - Fotolia
When Nidhi Gupta was selling logistics management software for DHL, she saw how large logistics companies were lagging behind when it comes to embracing cloud computing and automation.
“Only larger companies could afford our solutions, and at the back end we were dealing with static data and a lot of manual processes,” said Gupta. “Today, with machine learning, cloud computing and data science, it’s possible to do that in a more automated fashion.”
Sensing the opportunity to disrupt the market, Gupta started Portcast in 2017 with her co-founder, who has a doctorate in information engineering, to build a cloud-based platform that would make logistics more efficient and data-driven through the use of machine learning.
Portcast currently has 50 employees, including data scientists and software developers, across Singapore and Bangalore. In late 2018, the startup raised S$1m in seed funding and is in talks with current and new investors to secure a round of Series A funding.
Could you provide examples of how Portcast is solving some of the challenges that you saw at DHL?
Gupta: We predict the arrival times of cargo and the demand of cargo between any two ports. In terms of the arrival time, there are a lot of solutions from large software companies that do that, but the problem is they still rely on static data. Someone is manually feeding that data, or the data comes directly from shipping lines.
Now, that data is often neither complete nor resilient to external risks, like cyclones around the Panama Canal that may require ships to take a longer route or make a detour. We’ve seen ships being delayed by almost 24 hours before they get to the next port, which has a knock-on effect on every port after that.
Because we see weather patterns at a granular level, such as wave height, wind speed and cyclone intensity, we can combine that data with satellite data to predict if a ship is going to be late. We can quantify that, not just for the next port, but for the next five ports that the ship is scheduled to call at.
What that means is that the port, the logistics company and the shipper can plan their supply chain much better. They can avoid having trucks arrive at a port, only to wait for cargo. They can inform their factories and customers that the cargo is going to be late. Such a “just-in-time” supply chain is more resilient to risks. It’s like Amazon guaranteeing their delivery times.
Another example is unforeseen events, such as the port explosion in Beirut. We saw that ships that were supposed to go to Beirut on the day of the explosion were made to go directly to the next port. And when the ship arrives early at the next port, the shipping company will not have enough time to update customers that this is going to happen. What we can do is we can see where the ship is going, and the route it’s taking, to predict that it’s likely to arrive early.
The third thing is operational challenges. Shipping companies change their cargo movement all the time. We’re working with a large manufacturer that had cases where their shipping company decided to move 10 containers from one ship to another.
The containers arrived 10 days early, but there were no trucks because no notification was given to the manufacturer, which must bear the port charges for the extra containers until their trucks arrived. When you’re moving 200,000 containers a year, 10 containers might seem very small, but on a larger scale, the annual supply chain cost, especially during a downturn, really matters.
We also do demand forecasting – that is, to predict the cargo volume between any two ports. That allows us to figure out if a port’s capacity is sufficiently utilised, or if there is need for more, or less, capacity. Based on that, a cargo airline or shipping company can change their pricing dynamically, like how airlines price passenger tickets based on the number of empty seats.
You mentioned that other software suppliers rely on static data for their logistics management solutions. What’s stopping them pulling live data feeds to provide real-time updates, which is what you are doing right now?
Gupta: We use various datasets like weather data, satellite data about where ships are, data from shipping companies, data from ports and data on economic patterns. All that data is widely available, but the challenge is in orchestrating the data.
How do you get the data to speak the same language, maintain it, update it in real time, and use it to quantify the impact of weather and vessel movement on a ship’s arrival time? That requires machine learning that is customised to the logistics and supply chain space. The data is public, but how you orchestrate the data is a significant task and we’ve done that over two years, with almost four years of historical data now.
Larger software companies may have more resources, but their products are also widespread, and logistics may not be the only thing they focus on. That said, we’re discussing with some of the larger giants on how our technology can be integrated into theirs.
I presume most of your customers are shipping lines and logistics companies. Are there any other companies, such as large retailers, that might benefit from your technology?
Gupta: Yes, we started with the larger logistics companies and shipping companies. Amid the Covid-19 pandemic, we’ve received a lot of interest from manufacturers and large retailers that sell things like automotive parts and technology equipment – essentially, anyone who deals with containers. That market is more interesting because manufacturers and retailers want to have greater control of their supply chain. They don’t just want to depend on the logistics company to manage their supply chain.
Read more about supply chain management
- Having complete visibility over the supply chain will enable retailers to tide over the Covid-19 coronavirus pandemic and transform their businesses.
- Supply chain management seems to be changing profoundly as the coronavirus pandemic unfolds, possibly shifting from an economy of efficiency to resilience, harnessing trading networks, machine learning and robotic process automation.
- With analytical insights into supply chain management from IoT technology, organisations can better adapt to changing business conditions.
- Understanding which technologies can best simplify supply chain management is key to success in today’s economy. Here’s how three companies have tackled this issue.
You brought up the idea of a just-in-time supply chain. There have been suggestions that perhaps in the current pandemic and in the post-Covid-19 world, the just-in-time concept could also reduce supply chain resiliency. How can your technology help to mitigate that? Would your demand forecasting capability provide a counterbalance to the just-in-time supply chain?
Gupta: Absolutely. While Asia is the factory of the world today, there will be more balanced trade flows in future when more countries, such as those in Eastern Europe, could be production locations in a more diversified global supply chain. At the same time, we’ve seen many disruptions, not just Covid-19, but also cyclones and port explosions, that make supply chain resilience more important than before.
I think it’s not just about just-in-time; it’s also becoming more about “just-in-case”. But at the same time, supply chains can’t afford to have too much slack. So, the safety stock needs to be very optimised and therefore it’s more important to plan supply chains dynamically.
For one, it’s not enough to have safety stock that stays the same throughout the year. You have to plan for the entire year, and really look at demand and supply patterns, as well as the reliability of carriers – not just the cheapest ones.
These are the areas where we’re helping companies to figure out. Besides cost, which are the reliable carriers? Which ones are using greener routes from a sustainability perspective? That would offer a more complete picture when choosing shipping carriers.
As for demand forecasting, it’s about being able to plan up to a year ahead in terms of the kinds of booking patterns they will see and the capacity they need to procure. That helps in just-in-case or just-in-time planning.
Could you talk through the technology components of your software? How do you ensure your cloud service is resilient against public cloud outages from time to time?
Gupta: One of the key components of our technology is machine learning. We’ve built some proprietary algorithms from scratch and through co-creation with some of our biggest clients.
Another component is in the way we orchestrate data from multiple sources, manage and update the data in real time, and derive datasets from it. And through automation, we deploy data models to help customers make predictions almost at the click of a button for 20,000 different port-to-port pairs. That has become a very scalable process as it does not require any manual intervention.
In terms of making our service resilient, we’ve been through an extensive cyber security assessment process. We have the ISO27000 certification on information security and we do regular vulnerability scans to prevent anyone from hacking our systems. We also have cyber security insurance and follow all data protection guidelines.
In terms of cloud infrastructure, we work with reliable names like Amazon Web Services, Google and Microsoft, but we also make sure that we have business continuity and disaster recovery capabilities across multiple clouds to mitigate the impact of any disruption.
From a technology development perspective, what’s next in the pipeline? I presume you have APIs that your customers can use to integrate your software with their existing systems. Could you share more about that?
Gupta: Yes, we offer our technology through our website and through APIs which can integrate directly into our customers’ systems. There are two main things in our roadmap. The first is to expand our coverage to other shipping modes like air and road, beyond ocean shipping.
The second thing is we’ve been focusing on predictive AI, but we want to go into prescriptive AI. Now, we can say that the cargo is going to be delayed by 24 hours. We want to take things further and suggest alternative ships, routes or transportation options that could bring the cargo back on track.
For demand forecasting, we could say this is when demand is likely to be low, so you will have more empty assets around and now could be the time to change your pricing. Or, we can say these are the types of goods that customers are shipping more of right now. You could get your sales team to focus on these kinds of customers in a country. It’s about making our insights more actionable.