Colgate & Amazon Turned Recycling Data into Insight; Now Other CPGs Can, Too

“We enjoyed getting the information so much that we decided to expand to the second MRF in California,” Bedarf recalls. “The goal was gaining real-time insights into how tubes flow through a MRF.”

The data revealed that recyclable tubes were showing up more consistently in sort streams—even in communities that didn’t yet list them as accepted materials. “It showed us that the infrastructure is ready to handle tubes in more places than we realized,” says Greg Corra, VP of Global Packaging & Sustainability at Colgate-Palmolive.

“This is over time,” Corra says. “If you can show the data, everyone from MRF operators to municipalities becomes more comfortable saying, ‘Yes, this belongs in the bin.’”

The pilot replaced Colgate’s reliance on manual bale audits, giving the company continuous visibility into real recovery performance. “It’s a mutually beneficial data thing rather than a competition of who owns the data,” adds Brian McEnroe, who leads sustainability data analytics. “More data for these different entities is better.”

Over months of monitoring, Colgate saw a steady incline in tube recovery rates. Bedarf notes that each new deployment becomes easier as the AI learns. “You have to train it, and it builds up,” she says. “The second one’s a little easier, the third one a little easier. But it’s not like flipping a switch.”

For Colgate, DataStream isn’t just a measurement tool. It’s proof that design choices can translate to real-world outcomes, and that verified data can help drive local acceptance of new packaging materials.

A Glacier vision system flags a toothpaste tube among mixed plastics at a partner MRF. Colgate-Palmolive used this data to track real-world recovery of its recyclable HDPE tubes over time, replacing one-time bale audits with continuous monitoring.GlacierOther brand use cases

While Colgate’s project became the prototype for DataStream’s first deployment, Amazon was among the earliest companies to test how the same AI could answer a different question: what happens to bioplastics once they enter the recycling stream?

At the 2024 Packaging Recycling Summit, Glacier and Amazon presented data from a joint pilot in which AI systems identified bioplastic film and flexible packaging on a MRF paper line. Amazon funded part of the work through its Climate Pledge Fund, a $2 billion venture arm for low-carbon innovation.

“Back in 2019, we knew one of the key challenges was getting a handle on our boxes, plastic wrappers, and packaging,” said Nicholas Ellis, principal at Amazon’s Climate Pledge Fund, during that presentation. “We needed to understand what happens to it, then improve those designs to make them more recyclable.”

The AI model learned to detect small bioplastic items—produce bags, snack wrappers, and the like—with around 90% accuracy. The results showed that item size, color, and graphics can dramatically influence sortability. For Amazon, those findings informed packaging R&D decisions about material selection and labeling for improved recovery.

From these and other pilots, Glacier has already identified at least three core brand owner or CPG use cases for DataStream:

• Design validation — confirming whether packaging labeled “recyclable” is actually being recovered.
• Material optimization — understanding how design features like color or size affect sortation.
• Circular performance tracking — quantifying regional recovery rates and the effect of consumer education efforts.

“Brands come to us with different questions,” says Hu-Thrams. “Sometimes they want to verify that a change in resin or shape didn’t make packaging harder to recover. Other times, they’re looking for insights on how consumer behavior or labeling affects what makes it into the right bale.”

Hu-Thrams adds that brands beyond the initial pilots are already discovering new ways to use the data. “The technology is the same—it’s just the questions that change,” she says.

The data trust layer

Behind the analytics lies a simple model: Glacier sells access to its insights as a subscription service, built on data generated by its MRF-based AI systems.

Participating MRFs supply raw recycling data but retain ownership; Glacier aggregates it into anonymized, category-level insights. This approach gives CPGs a statistically robust picture of packaging performance nationally or regionally without revealing any single facility’s details.

For subscribers, DataStream functions like a typical SaaS platform. Clients log into a dashboard to view real-time recovery rates, trend graphs, and benchmarks. Subscriptions are modular: a brand may focus narrowly on a single format, like Colgate’s HDPE tubes, or scale up to monitor multiple categories. Glacier’s data scientists help interpret early results and integrate insights into packaging design or sustainability reporting.

Pricing tiers correspond to scope and complexity, allowing smaller brands to start small while providing multinationals with enterprise-level visibility.

For CPGs, the payoff is actionable insight; for MRFs, it’s operational transparency. The relationship is mutually reinforcing. The resulting data ecosystem is designed to reward openness, transparency, and continuous improvement.

AI detection distinguishes between toothpaste and non-toothpaste tubes, improving dataset accuracy and enabling brands to assess how different materials and graphics affect sortation efficiency.GlacierShared source of truth?

If Glacier’s early pilots of what would become DataStream proved anything, it’s that CPGs are eager for real-world feedback loops on packaging performance. With DataStream now active in MRFs serving one in ten Americans, the platform is building one of the most detailed recyclability datasets ever compiled.

For Colgate-Palmolive, the value lies in validation: proving that its HDPE tubes are being recovered in practice. For Amazon’s Climate Pledge, it’s diagnostic: learning which design attributes make flexible bioplastics more recoverable.

Both pilots helped shape DataStream’s foundation, but the brand outcomes speak loudest. Verified data has given Colgate new confidence in advocating for municipal acceptance of tubes, and it’s helping Amazon refine packaging design before products ever reach consumers.

Hu-Thrams sees this data as infrastructure for collaboration among brands, recyclers, and policymakers. “There’s a misconception that recycling is only a local issue,” she says. “But the truth is, the decisions brands make in design and labeling affect recovery everywhere. We’re helping connect those dots with evidence, not anecdotes.”

For packaging engineers and sustainability leads, that kind of evidence is fast becoming essential. As Extended Producer Responsibility (EPR) programs spread across states and countries, verified recyclability data will move from a “nice-to-have” to a requirement.

Hu-Thrams adds that circularity now depends as much on data as on design. “It’s not just about making packaging technically recyclable,” she says. “It’s about proving recovery and using those insights to make the next generation better.”