Introduction
Data analysts and senior management at a Fortune 500 entertainment company — and one of the largest streaming platforms globally — relied on multiple Tableau dashboards for content performance information and decision-making.
The Challenge
Optimizing visibility and performance when working with massive data
The client’s existing solution was born out of a compromise that sacrificed analytical capability for acceptable performance. In an earlier single-dashboard iteration, loading extracts would take hours and both extract and live query rendering would take 30-40 seconds for each filter interaction due to the size of the data set. To overcome these performance issues, our client built several dashboards to limit the underlying data, resulting in significant drawbacks:
- The dashboards lacked a single view of title performance, as they were limited to a single year of data and specific geographies, limiting our client’s ability to see long-term trends across multiple regions
- The multi-dashboard solution was difficult to maintain, with dashboard changes requiring modification of the 10-20 workbooks within the filtered data set
The Objective
A comprehensive and responsive single view of data
The client needed to combine the functionality and underlying data of the fragmented dashboard landscape to be able to assess the performance of original and non-original streaming titles alike across different regions in a single place, with KPIs such as %Reach, %Watch, %Watchers of Reach, and improve the response performance of the resulting dashboard.
The Solution
Optimizing Tableau using a proprietary toolset
Leveraging extensive Tableau experience and purpose-built inhouse tools, the HCLTech team discovered that Tableau was generating queries that had unnecessary joins in them, slowing queries in the database and transferring unnecessary data over the network. We:
- Applied several optimization methods, including checking the data sources and removing unnecessary joins, eliminating the unnecessary overhead from queries
- Pre-calculated the fixed values in a normalized table to further reduce the overall number of queries
- Redesigned the workbook data sources to leverage Tableau’s query fusion engine to serve multiple visualizations from single queries even if the views’ data aggregation levels were different
- Modified the data model to run queries faster and optimized how algorithms and KPIs are calculated using a custom extension to find hot and cold data within the database tables
The Impact
Faster and more impactful decision-making at reduced cost
The new, single-source-of-truth dashboard now holds over five years' worth of data and responds in a few seconds.
- Title managers can now access a comprehensive view of their shows' performance and drill down by a list of regions, times and a range of other dimensions, providing a thorough overview of title performance for hundreds of titles
- By eliminating three to four hours of Tableau Server background jobs, the client saves ~$30,000 annually on licensing and capacity
- The client saves approximately two FTEs worth of resources per month by maintaining only one dashboard that serves all use cases
