BigNoo - 173tech

Customer Behaviour

Granular product insights helped to fine-tune this virtual product

Objective

Understand user behaviour to optimise conversion and product discovery.

Obstacle

Lack of consolidated data on user engagement and product interactions.

Outcome

Product and category-level insights, enabling data-driven UX improvements.

Background

BigNoo is a fast-growing social commerce platform that blends community engagement with shopping, offering a curated experience that connects users through product discovery, sharing, and vaults, personalised collections of saved products.

As the platform scaled, the team wanted to gain a deeper understanding of how users were engaging with products, from browsing and adding items to vaults, to ultimately making purchases. By uncovering these behavioural patterns, BigNoo aimed to optimise its user experience and boost conversion rates.

To achieve this, the company partnered with 173tech to transform fragmented user data into meaningful insights.

Challenges

Fragmented Data Across the User Journey:User behaviour was being tracked across multiple tables and systems, making it difficult to form a unified picture of engagement. Metrics such as product views, vault interactions, and time spent were isolated from one another, complicating analysis and limiting visibility into the complete customer journey.

No Clear Visibility Into Product-Level Performance: While high-level site metrics such as total visits were available, there was no clear understanding of product-level performance. The team lacked insight into which items were most browsed, added to vaults, or converted most effectively. Without this visibility, it was impossible to make data-informed decisions around product curation, layout optimisation, or merchandising strategy.

Solution

Data Stack and Transformation: Our data stack is built around Google BigQuery as the central warehouse, with dbt used for transforming raw data and GitHub Actions for orchestration. Metabase serves as the visualisation tool, while source data is pulled from Matomo. Using dbt, we modelled this data into structured, analytics-ready tables. Core entities such as visits, actions, and visit actions were joined to form a central fact table, creating a consistent foundation for downstream analysis.

Behavioural Metrics and Visualisation: From the transformed dataset, we derived a suite of behavioural metrics to better understand user engagement and product interaction. These included total visits, average browsing time, top categories and products, most added-to-vault items, and most clicked products. We also calculated key conversion metrics, such as the visit-to-vault conversion rate and the number of vaults sent via email.

Impact

Improved Product Discovery: BigNoo now benefits from a unified and reliable source of truth for user behaviour, enabling the team to make more informed decisions around product placement and content relevance. By analysing which products are most frequently browsed and most commonly added to user vaults, the team can better understand what resonates with their audience. This insight allows them to prioritise high-interest products and categories, ensuring that users are consistently presented with more relevant and engaging content.

Enhanced Conversion Tracking and Iteration Speed: The introduction of detailed tracking for vault sends and user follow-through behaviour has unlocked a deeper understanding of customer intent and product desirability. By monitoring not just what users view, but also which products they save and later engage with, BigNoo can better interpret which items are likely to drive future action or purchases. This has introduced a new layer of behavioural insight, helping the team refine targeting and messaging strategies. Additionally, with self-serve dashboards powered by Metabase, cross-functional teams can now explore the data independently, validate hypotheses in real time, and make faster, data-backed decisions—eliminating bottlenecks previously caused by waiting for ad-hoc analysis.