How Uken Games Cut 100M Rows of Data Clutter

Uken Games Customer Story

“Using DataHub to kickstart spring (data) cleaning has delivered a lot of value for Uken so far.”

LISA STEPHENS
Data Scientist, Uken Games

CUSTOMER

Uken Games

INDUSTRY

Video Game Developer

SIZE

100+ employees

SOLUTION

DataHub Core (OSS)

USE CASE

Governance, Discovery

DATA STACK

Redshift, BI tools, ECS Fargate Spot tasks, ETL pipelines

GOALS

The Topline

Note: This story was originally published January 2023.

Challenge

Uken Games, a Toronto-based game studio developing titles including Solitaire Story: Ava’s Manor, Who Wants to Be a Millionaire?, and Jeopardy! World Tour, faced growing technical debt from years of ad hoc datasets and short-term, scrappy pipelines.

As Lisa Stephens, Data Scientist at Uken Games, explained: “We pull ad hoc data sets for one-off analyses, spin up short-term, scrappy data pipelines while more robust automation is in development, and more. We build up our own version of clutter.”

This technical debt led to several critical pain points:

Without a central source of truth, teams lacked answers to essential questions about cleanliness, granularity, time coverage, and context. This frustrated attempts at self-serve analytics and increased reliance on gatekeepers.

“Eventually, this creates a layer of technical debt … that dissuades people from self-serve analytics.”
Lisa Stephens
Data Scientist, Uken Games

Solution

Uken Games implemented DataHub as their central data catalog and governance platform, running their entire setup on ECS Fargate Spot tasks to optimize costs.

Their implementation focused on three key areas:

  1. Systematic Glossary Terms framework
    Using DataHub’s Glossary Terms feature, the team created a tagging system to capture and standardize metadata across datasets, including:
    • Quality level (bronze/silver/gold/iced)
    • Retention period (rolling date range, if applicable)
    • Granularity (row-per-X)
    • Data source (internal/third party)

This created filterable search capabilities and hierarchical term grouping, enabling users to easily answer questions like, “Where can I find a gold quality user-day dataset suitable for constructing a quarterly report about the in-game economy?”

  1. Query Statistics to identify low-usage data
    By leveraging DataHub’s usage data, the team identified low-usage data sets that were underutilized or abandoned. This insight guided systematic cleanup efforts, reducing storage costs and minimizing the risk of decisions based on stale or incorrect data.

“When we recently used this feature to evaluate a family of tables that had historically been used by our data science and product management teams, we found that as many as 40% of them, amounting to approximately 100 million rows, were no longer needed.”
Lisa Stephens
Data Scientist, Uken Games

  1. Custom Metadata to link out to existing documentation
    Uken used DataHub’s entity-level About sections to link out to existing documentation, allowing end users to more easily find the missing context they’re looking for. Depending on the nature of the table in question, the team includes:
    • A brief summary of what the table contains
    • If the table comes in variants of different cleanliness or granularity, references to what those variants are, and a summary of how they differ
    • If the table has been deprecated, a reference to the newer version that should be used
    • Hyperlinks out to the original spec, analysis, or monitoring dashboard
    • Hyperlinks out to the corresponding ETL job

Impact

With DataHub, Uken Games built a cleaner, more navigable, and self-serve-friendly data ecosystem.

Key outcomes included: