Waiting for perfect data delays progress indefinitely. Maritime operators can extract value from imperfect data today while improving quality over time through better workflows, validation at entry, and user-centered design. The key insight: data quality improves through use, not through waiting.
Data quality is often cited as the primary obstacle to AI adoption in maritime. The concern is legitimate—garbage in, garbage out remains as true for machine learning as it was for earlier generations of analytics. But treating data quality as a prerequisite rather than a starting point has stalled more initiatives than bad data ever has.
"I'm going to say the boring one first: the data quality. As everyone knows, that's always an issue if you want to do whatever you want to do in digital, not only AI. Even if it's 2025, we still struggle a bit with the data from the ships still."
— Nils Israelsson, Head of Operations, Stena Bulk
The challenge is real. But the response to it—waiting until data is perfect before attempting anything—is a trap. Perfect data does not exist in operational environments, and the pursuit of it can become an indefinite delay.
The Seven-System Problem
To understand why data quality is so persistent a challenge in maritime, consider how data actually flows on board a typical vessel.
"We have captains on, we have different pools, et cetera, so we see different ships from different owners and charters as well. People manually fill in the daily remaining bunkers on board, so they fill that out in seven different places—they need to send it to charters, to this system, to that system. Of course you're going to have data quality issues."
— Nils Israelsson, Head of Operations, Stena Bulk
This is not a technology problem that technology alone can solve. It is a workflow problem. The same information is entered multiple times into disconnected systems, each with different formats and validation rules. Errors compound. Inconsistencies accumulate. And the people entering the data—already stretched thin—have no visibility into how it will be used downstream.
Data Quality Starts with the User
The insight that separates effective data strategies from ineffective ones is recognizing that data quality is fundamentally a user experience problem.
"Data quality starts from the user and we're currently asking the user to have double, triple time and enter the same data on different systems. We need to always go back to the user who's entering the data and the other user at the other end who'll make decisions out of this data."
— Alexandre Lapointe, Chief Product Officer, OrbitMI
This means designing systems that reduce duplication, validate data at the point of entry, and give users meaningful feedback—not just error messages that say 'speed must be positive' but real context about whether the data makes sense.
It also means thinking carefully about what data is actually needed. Not everything must be precise. Some decisions require exactness; others work fine with approximations.
"When you look at optimization, sometimes ish is good enough to take some commercial decisions."
— Alexandre Lapointe, Chief Product Officer, OrbitMI
Garbage In, Garbage Out—But Start Anyway
The foundational principle remains valid: poor inputs produce poor outputs.
"If your data is not on par, start there. Because if the data is garbage, garbage in, garbage out. That's a clear point."
— Felix Jan van den Bos, Independent Digital Transformation Consultant
But 'start there' does not mean 'wait there.' It means beginning to address data quality while simultaneously exploring what can be done with current data. AI does not require pristine datasets to deliver value. Directional insights, anomaly detection, and pattern recognition are often possible with imperfect information.
The key is matching ambition to data reality. Some use cases demand high accuracy. Others tolerate uncertainty. Understanding which is which prevents both paralysis and disappointment.
Quality Improves Through Use
Perhaps counterintuitively, one of the most effective ways to improve data quality is to start using data for decisions that matter. When data has visible consequences, attention to quality increases. When feedback loops are tight, errors get caught and corrected faster.
This is why the best data quality initiatives are not standalone cleanup projects but integrated parts of operational improvement. They connect the people generating data to the people using it, creating accountability that no validation rule can replicate.
This series is based on a webinar from Digital Ship. Watch the video here
.png)
