Thinklytics

AT&T · Technology & SaaS · Dallas, TX · 16 weeks

AT&T Analytics & BI case study

AT&T faced serious server slowdowns with 6,000 Tableau workbooks built over seven years, causing executives to abandon dashboards before they fully loaded. Instead of spending $6.2 million on a platform migration, Thinklytics optimized the existing setup and restored performance.

Challenge

The Tableau Server had ballooned to 6,000 workbooks, with most unused for over 18 months. Dashboards took 47 seconds to load on average. The vendor recommended a complete BI platform migration, costing $6.2 million and taking 18 months.

Approach

We used Tableau’s REST API to gather data on workbook usage, last-accessed dates, and extract sizes. We found 4,380 workbooks with no views in the past year. After notifying stakeholders for 30 days, we retired those unused workbooks. Next, we rebuilt the 40 most-used executive dashboards with optimized extracts to improve performance.

Outcome

We cut dashboard load times from 47 seconds to 9 seconds. We retired 4,380 outdated workbooks, which made the planned $6.2M migration unnecessary. The remaining 1,620 workbooks now run efficiently on a server operating at just over a fifth of its capacity, improving overall system performance.

The one thing we overlooked in the migration plan, and how it caught us off guard

The vendor kept blaming the platform, but frankly, the real issue was seven years’ worth of workbooks stacking up on an untuned server. So, we started by digging into the usage data to get the full picture before making any changes.

We retired 4,380 unused workbooks without disrupting any active users.

We focused on the 30-day notification window. We sent every workbook owner an email with a direct link to their workbook and a clear deadline to ask for it to be saved. Out of 4,380 workbooks scheduled to retire, only 12 got saved because the owners reached out.

We tackled 40 workbooks head-on, fixing data glitches and speeding up reporting big time.

We tackled 40 key executive dashboards and gave them a solid tune-up. Cleaned up the data pulls and set up automatic refreshes. The payoff? Load times plummeted from 52 seconds to just 7. After that, the C-suite ditched their printed reports and jumped right into the dashboards.

Results

  • 4,380 Workbooks retired
  • 47s to 9s Dashboard load time
  • $6.2M Platform migration avoided
  • 22 of 100 servers retained Server capacity after cleanup

We were three weeks from closing a $6 million migration deal when Thinklytics showed us the usage data. That’s when we realized we were about to spend $6 million fixing a problem we had mostly created ourselves.

Director of Analytics Engineering, AT&T

In the client's terms

The question

What decisions is this reporting supposed to support?

The situation

Tableau Server held 6,000 workbooks, most untouched for over 18 months, with dashboards averaging 47 seconds to load. The vendor recommendation was a full platform migration at $6.2M over 18 months.

What we found

The server was not slow because the platform was wrong. It was slow because it was carrying several thousand workbooks nobody had opened in a year. Migrating all of that to a new platform would have moved the problem at considerable expense.

What it produced

Load times fell from 47 seconds to 9. The 4,380 retired workbooks removed the case for the $6.2M migration. The remaining 1,620 run on a server at just over a fifth of capacity.

The roadmap

Usage, last-accessed dates and extract sizes came out of the Tableau REST API rather than from asking people what they used. That identified 4,380 workbooks with no views in twelve months. A 30-day stakeholder notice ran before anything was retired. The forty most-used executive dashboards were then rebuilt on optimised extracts.

The work behind this result

Related case studies

Frequently asked questions

Is a slow BI server a reason to migrate platforms?

Not usually. AT&T's Tableau Server was averaging 47 seconds to load and the vendor recommendation was a $6.2M platform migration over 18 months. The actual cause was 4,380 workbooks with no views in twelve months. Retiring them took load times to 9 seconds and removed the case for the migration.

How do you decide which dashboards to retire?

From usage data, not from asking. At AT&T, workbook usage, last-accessed dates and extract sizes came out of the Tableau REST API, which identified 4,380 workbooks nobody had opened in a year. Stakeholders then had 30 days to object before anything was retired.

What does BI sprawl cost?

At AT&T it was the case for a $6.2M platform migration that turned out not to be needed, plus 47-second dashboard loads across the business. After rationalisation the remaining 1,620 workbooks run on a server at just over a fifth of capacity.