Why data projects fail in maritime
02 Apr 2019 - Nick Chubb
Data has been hailed as the technological saviour of maritime, with the potential to solve just about every problem shipping faces. The current reality however, is that a lot of data analytics projects fail, sometimes spectacularly, a problem that appears to be particularly prevalent in maritime. Here are three reasons why your upcoming data analytics project might fail, and what to do about them.
You’re asking the wrong questions
There is a misconception that exists in our industry about what data science is and how it should benefit us. First, we start collecting data, then we hire some data scientists or get a consultancy in, then we will enter a zen-like state where the machines make our decisions for us while we sunbathe. The unfortunate reality is that without asking the right questions first, any attempt to become data-driven will fail, usually taking hundreds of thousands of dollars with it. This is a problem that exists across every enterprise business but is particularly prominent in maritime, principally because it is difficult and expensive to collect and process data from ships making those inevitable failures more expensive.
The most important thing to get right when commencing a data analytics project is to ask the right questions. With enough of the right data, you can answer any question but if you don’t go in with a clear idea of what you need to know you will find out nothing, or potentially find out a lot about something irrelevant.
You’re over-complicating data collection
Collecting data can be expensive in this industry. IoT is the latest buzz area when it comes to data collection but fitting sensors that will work with little maintenance in a maritime environment is not easy, and once those sensors are fitted getting the data off the ship in real time can be an expensive undertaking.
As well as the complexity and expense, there are contractual issues to contend with too. Where is the incentive for a ship owner to fit IoT sensors to a vessel if the net benefit will come in fuel savings? The owner will have to outlay the expense and the charterer will get the saving. It is unlikely that an owner will see a return for fitting sensors as the market is volatile and charging a higher spot rate could wipe out any fuel savings for the charterer.
Often the simplest way to begin data collection work is to look at what is currently available. Huge amounts of data is generated across a ship every hour without the need for additional sensors. At Intelligent Cargo Systems we investigated the viability of a number of different ways to collect container terminal performance data in real time and eventually realised that the crew were already logging everything we needed in paper logbooks. Once we digitised the log book, we had access to the data that already existed we could use it to model and predict terminal performance in ports worldwide. Rather than install fuel consumption sensors to a ship, maritime startup We4Sea uses publicly available data sources including AIS, weather and tides, and engine specifications to calculate the fuel consumption and emissions of any ship in real time. Before committing to installing new hardware on a ship, be sure to audit what data you currently have access to, you will be surprised what can be achieved.
No one cares
Data without insight is worthless and insight without action is pointless. Assuming you are asking the right questions and are generating useful answers to those questions, how do you turn what you are learning into action?
The simple truth is that it is incredibly difficult to change behaviour in large organisations. This is the point where a technology project becomes a change management project. If you don’t have the buy-in of all of the stakeholders required (both internal and external) to bring about real change, you will fail. At an individual level, while there may well be a desire to improve, the day to day pressure of the job will often get in the way.
At a higher level there is often a disconnect between what a technologist thinks is valuable and what an end user will actually find valuable. Putting the right information in front of the right people at the right time and in the right way is as much an art as it is a science and requires an incredibly close collaboration between developers and end users. Humans respond to stories, not data. If you want people to take action, you not only need them to care but you need to present them with data in a way that, as a minimum, fits with their current workflow, and in an ideal world provides a drastic improvement.
Lastly, where new and emerging technologies are concerned, the only guarantee is failure on some level. The biggest lessons come from those small failures and it is important to keep a project alive long enough to learn from them. If you are starting a maritime data project you should expect small failures, but if you ask the right questions, think carefully about how you collect data, and ensure that everyone is invested in the outcome you will avoid a much bigger failure.