Quality not quantity

How clean is your data? Can it be trusted? Is optimisation being achieved proactively or is it always a reactive function with analysis after the fact? Do you resort to spreadsheets in an attempt to reconcile data glitches?

Chad Van Derrick, Vice President of Software Product Management at Tideworks Technology, says that every single meeting he attends with ports and terminals is peppered with talk of optimisation: “But what does it actually mean? It is really about data and the integrity of the data, and making sure it is actionable and easily accessible.

“It can be as detailed as attributes around a single container – we were expecting this, it isn’t there – or very subtle details. But if the data is wrong, then it’s a case of death by a thousand cuts. Enough of those errors and all of a sudden, you are not planning as efficiently or correctly as you should, and a container gets released from Customs when it shouldn’t be or a hazardous container isn’t flagged up.”

Tideworks provides end-to-end software for marine and rail terminals, from documentation and billing to the planning and execution of vessels, as well as the data platform for analytics, AI and machine learning.

Van Derrick says that when operators are looking into or adopting new systems, data is often ‘lost in the shuffle’, becoming an issue almost at the last minute, particularly in system migration. “It’s often ‘wait a minute, we should look at this’. When system migrations don’t get adopted, it’s often because the data can’t be trusted. The question is, why not?”

CRUCIAL FACTOR
Data governance is crucial for shaping the future of marine terminal operations and enhancing the value and management of near real-time data, he says. Governance in this context includes compliance and data security, but most importantly it is about policies and procedures.

“Data is typically broken down into three basic areas – master data, transactional and analytical. That requires different governance along its journey because things can change. What’s in one system might conflict with another, which might conflict with another. So, it’s about knowing the data journey, where it is stored, etc.”

This requires different policies to ensure integrity, he explains. “At its simplest – if a particular value shouldn’t be negative or null but should be within a certain range and is not, the system should set up a flag because something’s not right and requires someone to take a look and possibly clean it up.

“Instead, a lot of the time there are teams of folks operating on spreadsheets and email to try to reconcile the data. It takes a ton of effort, intervention and paper processes. But with the correct data governance, the system can flag it up, and a data steward goes in to make that adjustment. It is about picking up erroneous stuff early – the longer those discrepancies last, the more probability of error that can have a really significant impact in our industry.”

It’s important to build the foundations from the start, he says: what is the master data, make sure it remains clean and high integrity, with boundaries around it so that this can be maintained.

“The goal is to change from people manually going in and resolving issues after the fact, to being proactive and providing insights early on. That is the golden key to optimisation.”

Tideworks is working with a number of terminals on AI and ML projects to deliver optimisation. An example of this might be an ML model to predict dwell time, based on the last three months of data. “The model is only as good as what you are putting into it. Any data anomalies contribute to the model – as you then try to automate the model and get into AI, you are moving containers and planning the stacking based on what you think is the dwell time. If it’s wrong, you will be putting the containers in the wrong spots or moving a container to another position where another container is already sited, so wasting fuel and labour and resulting in chaos and maybe even safety issues.”

Questions must be asked around any data feeding into such a system: has it been cleansed, standardised, enhanced? Are we comparing apples with apples? Different terminals measure various metrics in different ways.

“That’s super important when you get into ML and AI. If we don’t understand the data being fed into those models, the results will be interpreted wrongly. AI will take the data for what it is and look for correlations.”

Overnight terminal housekeeping really benefits from clean data and data governance, he says. “Decisions can be made to put this container on top of this one, move another, optimise the yards. Also, data can tell you what the gate turnround times are, where trucks are getting backed up, where the waits are, so that you can redeploy resources accordingly. Just making changes in operations based on what you are physically seeing can have other downstream effects.”

Are ports making the most of the opportunities from data and real-time analysis? Van Derrick says some terminals are close, their aspiration is real-time, and they are perhaps getting within 30 minutes of real-time in terms of getting access to data and making decisions based on it. “Other terminals have big monitors above their operations desks which are just blank – they are not tracking what I would deem to be essential KPPIs to run their operations, but looking at KPIs after the fact. Data can be retroactive, looking at performance and taking out insights for the following week. But that’s what we need to shift to get into the new optimisation – pulling in data real-time and acting more proactively.”

UNDERSTANDING GOALS
Terminals should start by understanding their business goals, whether it’s increasing capacity, throughput or handling times, for example. “Optimisation in itself isn’t a business goal. Define what you are trying to achieve and technologists can look at how to help make it happen. Defining the ‘what and why’ is so critical and yet it’s amazing how that’s the difficult thing for many terminals to do. Sometimes the terminals are so in the weeds day to day that it’s hard to take a step back and consider how to increase frequency or profit.”

Data for the sake of data is certainly a pitfall, he says. “You can get so wrapped up creating this huge data lake and trying to optimise it and basically look for a nail that isn’t there. You can have the biggest hammer but if you don’t have the nail and you are finding that after the fact, then maybe you didn’t need the hammer.

“I see a lot of terminals go down that path. But if they can’t articulate what they need and what they are trying to achieve, then I have to ask the question – what is all this effort going into? You can throw a lot of time and money at a project but the initiative is unlikely to work.”

Other pitfalls are thinking that data governance is a one-time event and thinking that everything must be done at once. “Often, data governance work is tied into a data migration project, into ERP, TOS, etc. – a sort of ‘time to clean up the data’. That’s all well and good but it can be a very overwhelming one-time event. You throw resources at it, suddenly you are in the new system and it all ‘goes away’, but a year later, you are in the same position. It’s really important to establish this in bite-size amounts and continue to come back and ask – are we will doing the right thing? Is the benefit still there?”

The amount of data itself can be an overwhelming burden. He advises: “Take time to identify the top 20 pieces of data that if they’re wrong are really going to mess things up and affect operations, safety, financials, for example. Draw a box around these; focus on these areas and see the results of that effort.”

Where will the industry be in five years’ time? Data will become increasingly important, says Van Derrick. “I would like to say we will be exchanging data between us, across terminals, in a more governed, standardised and readily available state. I think we will be getting there but will not be here yet.

“I think we will see some experiments in AI – some may be successful, some maybe not so much. But we will see more business rules and machine learning come into the mix, with things like housekeeping being automated and more predictive applications for tricker tasks. With the likes of ChatGPT entering the consciousness, terminals have the interest to see what they can do with that. It will bring up some really interesting fundamental questions and problems. The next five to ten years will be really pivotal for the industry.”