DUCK Transforms Customer Analysis with Neo4j and LLM-Powered Netnography Platform

Most companies miss vital signals from their customers. “It’s flabbergasting,” says Justo Boero Delbanco, CEO and co-founder of DUCK, a Swedish technology agency that works with the world’s largest and most recognizable brands. “All these people are talking about you, talking about your competitors, and you spend endless hours trying to analyze the signal behind the feedback.”

DUCK was founded to transform the way brands connect with their customers. Even as companies collect mountains of feedback across product reviews, social media, and customer service logs, they struggle to make sense of it all. DUCK’s proprietary AI platform, Kiku, connects these disparate customer datapoints into a Neo4j knowledge graph that can be queried using natural language — giving brands a powerful tool to cluster, assess, and react to trends in real time.

This analysis approach is known as netnography. This research method seeks to understand online communities and examine how people form identities on platforms like Facebook, Reddit and YouTube.

“A relational database is good for storing data like social media posts, but it can’t provide the revelations we get by illuminating connections in a data graph,” says Henrik Johansson, DUCK’s CTO and co-founder. This architecture allows DUCK to cut through the complexity of human feedback, finding patterns that would be missed in traditional databases.

Breaking Through the Noise in Customer Feedback

DUCK began its graph journey by helping clients decode the complexity in their product reviews.

“A five-star review for an apparel company could be solely about the purchase experience, and not the product itself. And in a three-star review, the product might be perfect. Maybe the fit and color were ideal, but the purchase experience was terrible,” says Delbanco. “Neo4j graph technology helped us cut through that complexity to get to the key insights. It offered richer contextual analytics behind the reports and dashboards our clients need.”

Neo4j’s Cypher query language helped DUCK adapt quickly as they discovered new patterns in customer feedback. The team could add new attributes and relationships without rewriting their database structure – essential when analyzing everything from product reviews to social media conversations. And when every hour of developer time costs thousands in overhead and delayed market opportunities, this flexibility meant DUCK could invest in creating value for customers.

Eliminating GenAI Hallucinations with a Neo4j Graph

DUCK’s next breakthrough came through their work with Dr. Marcus Collins, former head of strategy at Wieden + Kennedy and marketing professor at the University of Michigan. As Collins prepared to release the paperback version of his book For the Culture, he wanted a way to share his insights about cultural marketing beyond the page.

The result was ChatMTC. Accessible via a QR code on the book’s cover, this GenAI assistant answers questions about how culture influences human behavior and shapes marketing. DUCK built the assistant using Collins’ complete body of work – from his doctoral thesis to his latest articles and lectures.

“Getting both the tone and facts right was critical. Any mistake could damage Collins’ personal brand,” says Johansson. The solution lay in Neo4j’s knowledge graph. “We use the graph as our foundation of facts, and LLMs for what they’re best at – reasoning with that data. Most AI hallucinations happen because people use LLMs as knowledge libraries, which they were never designed to be.”

Today, DUCK builds these knowledge foundations using AWS and Amazon Bedrock with Neo4j. The team chose a middle path between traditional Retrieval-Augmented Generation (RAG) and GraphRAG approaches. “We focus on functional techniques like parental retrieval that deliver real results,” says Johansson. “We appreciate that Amazon Bedrock is serverless so that we can quickly set up new proofs of concept to make sure the algorithms, queries, and results work as they should,” Johansson says.

ChatMTC’s success showed DUCK how to turn their approach into a platform any company could use. Building on this foundation, the team developed Kiku, their netnography platform specifically designed for brands that truly want to interact with their customers and to do that in a authentic way to understand the person behind the customer.

“With Kiku, we’re not just analyzing what customers are saying, but understanding the context, connections, and cultural significance behind those conversations,” explains Delbanco.

Building the world’s largest consumer intelligence graph company

Over time DUCK has expanded its use of Neo4j capabilities, leaning on the Neo4j Graph Data Science (GDS) Library for techniques to embed algorithms on the graph itself, such as with the Fast Random Projection (FastRP) algorithm. The FastRP algorithm allows DUCK to efficiently analyze relationships between millions of data points to surface hidden patterns that would be impossible to detect manually — critical for identifying emerging customer trends and sentiment shifts.

“As we iterate our solutions like Kiku, we often layer in different LLMs and embedded algorithms to produce better results. The traversal and retrieval speeds we need to handle this complexity would not be possible without Neo4j,” says Johansson.

By integrating data from sources like Reddit and public Facebook posts — or any other source where people talk free and willingly and express their opinion — Kiku provides brands with a self service tool where users can interact with the Intelligence graph in a conversational matter and based on the question. Kiku collates online conversations into specific topics and trends — exposing the underlying motivations and values driving consumer behavior, “To make the intangible tangible,” says Delbanco.

For example, a fast food chain might tie bubble tea sales back to a surge in interest for East Asian culture, while a car manufacturer could monitor emerging trends around specific car meets or enthusiast clubs in particular cities, or identify influential voices in their category. 

Kiku allows different teams to extract targeted insights from the same data, each from their own unique perspective. A marketing team can apply a cultural lens. An operational team can explore the same data using a sustainability lens. Integration with GenAI then allows teams to query trends with natural language to develop new strategies, tactical initiatives and operational actions. This crucial step makes the data actionable over the entire organisation, where it creates value.

Above: Watch the Neo4j Connections Virtual Event with Justo Delbanco, CEO, Duck and Andreas Kollegger, Senior Developer Advocate, Neo4j

Get in Touch

Curious about what insights you could unlock for your business with graph-powered solutions? Let’s talk — reach out, and we’ll get in touch.