Most companies don’t have a data problem. They have a data chaos problem.
AI promises game-changing insights, but without a solid data foundation, businesses might find themselves layering automation on top of disorder. The real challenge isn’t collecting more data, it’s making sense of what’s already there.
That means cleaning it up, structuring it properly, and ensuring people actually trust it.
At YYC DataCon 2025, three experts from Arcurve broke down what it takes to escape the “data swamp” — a growing problem where valuable information gets lost in an unstructured mess.
Eric Nosal, a technology principal with decades of experience in big data and machine learning, explained why flashy AI tools won’t fix a broken data strategy. Software developer Taylor Noel has spent enough time fixing broken reporting systems to recognize the patterns that lead to inefficiency. And Tony Truong, a delivery partner focused on data governance, has seen firsthand how businesses rush into AI without first building the right foundation.
To put it bluntly, Nosal said, “Data strategy isn’t about data. It’s about the business value and the plan.”
Here’s how to stop drowning in data and start making it work for you.
Step 1: Identify and eliminate the “band-aid” fixes
Many organizations layer quick fixes onto existing data workflows until the entire system becomes unmanageable.
Noel shared a case where a client’s reporting system had grown to 10,000 views per month but was full of errors and inefficiencies.
“There were always bugs that came in, there were fixes, there were new requirements that needed to be addressed,” she said. “They were addressed kind of on a one-off, so they had band-aids on top of band-aids.”
To move from a fragile system to a reliable one, businesses should:
- Map out workflows: Identify where errors and inefficiencies occur.
- Engage business users: Sit down with employees who rely on the reports to understand what’s actually broken.
- Standardize definitions: Avoid situations where key metrics, like “cost of goods sold,” are calculated multiple ways.
- Automate repeatable tasks: In one case, an employee was manually extracting and uploading data every day, introducing errors and inefficiencies. Noel’s team identified this as a repeatable process and automated it, saving time and improving accuracy.
“By identifying that as a repeatable process, we were able to automate the process itself,” she said.
Step 2: Stop throwing everything into a data lake (or worse, a data swamp)
A common pitfall is assuming that dumping all company data into a central repository — a data lake, if you will — magically leads to insights. Nosal warned that without structure, this quickly turns into a “data swamp.”
On stage, Nosal described a situation where a company’s data was flowing into a central storage system, but without clear organization, much of it was unstructured and difficult to use.
“In this particular case, we didn’t have a data swamp. What we had was maybe more like a data lake with a swamp kind of forming around the fringes,” he explained.
One of the biggest culprits was PDFs.
“For us data people, it’s one of the biggest nightmares,” Nosal said.
Why? PDFs often contain inconsistent formatting, images, and handwritten scans, making it nearly impossible to extract clean, structured data. The company in this case had a wealth of information locked inside PDFs, but without proper structuring, the data remained difficult to analyze.
To fix this, Nosal’s team implemented automated ingestion pipelines to process and structure the unstructured data.
The process involved:
- Extracting text and tables from PDFs using AI and machine learning tools.
- Standardizing formats so data could be queried and compared.
- Eliminating manual processes that slowed down data access.
Once the data was structured, employees could finally use it to drive real business decisions.
Although this remains a challenge in the industry, using structured ingestion strategies has made previously inaccessible data more usable for business insights.
Step 3: Build trust in data through automation and accuracy
Even the most advanced analytics tools could be rendered useless if employees don’t trust the data. Errors, inconsistencies, and unreliable reporting can lead to skepticism, making it harder for teams to adopt data-driven decision-making.
One of the biggest issues Arcurve encountered was the impact of manual processes introducing errors.
Noel explained that the team used DBT (Data Build Tool) to implement automated data quality checks, helping ensure data accuracy by applying validation rules and identifying inconsistencies. If critical sensor data was missing, the system would automatically flag those records and return them to the source for correction.
By automating validation and flagging missing or incorrect data, they improved accuracy while reducing the need for employees to manually correct reports.
So how do you do it?
- Minimize manual work that introduces human error. Identify and automate repetitive data-handling tasks to ensure consistency.
- Implement automated data quality checks. Use validation tools to flag missing, duplicated, or incorrect data at the source.
- Ensure transparency in data processes. Employees need to understand how data is structured and where it comes from before they can fully trust it.
- Make reporting accessible and real-time. The faster teams can access accurate data, the more confident they will be in using it for decision-making.
By addressing these issues upfront, organizations increase confidence in reporting, reduce inefficiencies, and create a culture where data is seen as a trusted asset rather than a liability.
Step 4: Measure ROI the right way
For executives asking, “How do we prove the ROI of our data strategy?” Truong laid out three levels of value measurement:
- Direct ROI: The easiest to quantify. If a company saves 20 hours a week by automating a report, the cost savings are immediate.
- Inferred savings: Automating data pipelines means reports can run more frequently, leading to faster decision-making.
- Risk reduction: What’s the cost of bad data? Truong explained that poor data quality could lead to costly mistakes — such as fines due to incorrect compliance reporting or operational inefficiencies.
If your organization is struggling to see the ROI from AI, take a step back. Start with your data.
“The easiest one for us is automation,” said Truong. “Some organizations do a bunch of manual processes to create a report, so that ROI is super simple, right? You automate that process and you save that person’s time, so however many hours you spend per week or per month building that report, that translates to a direct ROI savings.”
The bottom line? Data strategy is an investment in business stability. Organizations that measure ROI across efficiency, risk, and decision-making will see long-term value, not just short-term savings.
The takeaway: Data transformation is a journey, not a switch
Organizations often ask if they can reach AI-driven insights within months. Truong’s response?
“Well, yes and no, because the organization needs time to understand the data, understand the insights they have so they can bake it in and ask them questions.”
Instead of treating AI adoption as a quick win, the panel emphasized that companies should focus on long-term data maturity.
“It’s really 80% people, 20% technology,” Nosal said. “That’s something that we try to try to live by.”
Companies that take the time to clean up their data infrastructure, automate workflows, and build trust in reporting will be far ahead of those blindly chasing AI trends. AI can be a powerful tool, but only if the data it runs on is organized, trusted, and accessible.
Digital Journal is the official media partner of YYC DataCon 2025.
