Mattan Benyamini

Data Analyst Team Lead at Windward
Windward logo

Windward is an international predictive intelligence company focusing on maritime traffic. Our platform serves as a one-stop-shop for all maritime domain needs, running risks for vessels and so on and so forth. 

The Windward solution fuses AI, big data and maritime expertise to enable clients and partners to understand the maritime ecosystem and its broader impact on security, finance and business – and by doing so allows them to make data-driven decisions. 

Commercial and governmental clients come to us and we help them analyze and produce insights regarding their vessels, by analyzing multiple data points, which we feed into our sophisticated artificial intelligence models.  

Our commercial clients include insurance agents that want to know the risks of their vessels in the context of accidents and maritime casualties. We compile this information by looking through banks that fund deals between different commercial entities, and most recently we expanded our offerings to provide ocean freight visibility to freight forwarders and other cargo owners, which helps to predict the ETA of when their goods will arrive at port.

For government entities, we perform “border security risk” assessments, in which we identify vessels that are not operating in any economic capacity, to point out suspicious carriers for these governing bodies to keep an eye on – allowing them to better protect their waters and borders from maritime threats.

Looking at our Ocean Freight Visibility product that we just launched. We’re implementing state of the art technology to solve a really complex problem that the world has been dealing with for a long time, and more recently this problem has been at the core of multiple market crises. The problem in predicting ETA for cargo vessels and containers is that a lot of aspects have to do with the arrival time of a cargo vessel and a container. 

The technologies that have been around in the past decade are decent, but not good enough for this kind of problem. This is because the technology needed to consume the layers of data, and public web data, that are involved in bringing full visibility to maritime activities was not yet available to the market – but now it is.

Windward chose a technology called Deep Learning to power our platform. So, we use a neural network that essentially knows how to interact with the different data sources and combine them to reach one conclusion. In this case, the estimated time of arrival of the vessel at the port destination.

Focusing solely on the ETA model, because the entirety of our operation is quite complex. The basic layer data that we’re using is the transmissions of vessels that constantly send signals to various receivers all over the world, every minute, to keep track of the vessel’s whereabouts.

We use these different vendors to map out the location of these vessels using the transmission data they provide to us by the minute. In the scope of things, we are typically looking at hundreds of millions of vessel transmissions a day.

But when we try to predict the time of arrival of the vessel to a port, we need to consider different data sources, such as public web data, and one of the most important web data sources for this is the vessel schedules posted along the different shipping carriers’ websites. This information includes the last known location of the vessel, its current whereabouts as well as the estimated time of arrival at port.

Open-source web data is so important to Windward because we are using those carrier websites to feed our algorithms in order to automatically predict those ETAs, and help companies focus on other aspects of their operations.

You can imagine a container ship as a bus, and the bus collects people from different stations. Now, imagine every person in the bus has his or her own prediction of when they will arrive at the destination. In this analogy, the people are the carrier websites. 

So, it’s not enough to ask just one person when they think they will arrive. We need to ask multiple people and then average them out into insights. Therefore, it’s important to use different web data sources on the internet, and not just one. 

To collect the public web data that feeds our algorithms, we use Bright Data’s Web Scraper IDE to automatically pull web data from the different shipping carrier websites, and we have been using this solution for a few months now.

During the brief time we’ve worked together, I think what we have accomplished has been super productive and conducted with quick execution – formally addressing our needs as a company.

Furthermore, if we ever hit a snag in our web data collection, Bright Data’s support staff was also responsive, and aware of how to solve the issues, fixing it immediately or within a reasonable timeframe.

So, to me, personally, I feel the trust working with Bright Data.

More testimonials

Head of IT, ClickBus
We use public web data to understand the competitive dynamics of a market. In digitized markets, we can get all the offers capturing the web information.
Timo Helken
CTO at Remazing GmbH
We use public web data to understand the competitive dynamics of a market. In digitized markets, we can get all the offers capturing the web information.
Felipe Bianchi
COO, Gregario
We use public web data to understand the competitive dynamics of a market. In digitized markets, we can get all the offers capturing the web information.