Case Study · NBA League Pass · 2014 to 2021

The role was invented because the customer experience made the news

League Pass had a customer satisfaction problem fans could see. Support was overwhelmed, the Better Business Bureau rated the service D, and each season opened with another round of bad press. As the 2013-14 season opened, Deadspin ran two pieces in thirteen days telling fans not to buy. Turner Sports and NBA Digital created a role to fix it. I was hired into it midway through that same season, in February 2014, and I ran customer satisfaction as a product until 2021.

Read the seven steps Skip to the result
16Platforms supported, web included
370k+Subscribers served
4×Subscriber growth over four years
DABBB rating, within two years

How it was built

Seven steps from a public complaint to a measured practice.

1The call

A product experience so bad it reached the sports press

As the 2013-14 season opened, Deadspin ran two pieces in thirteen days telling readers not to buy NBA League Pass Broadband. Support was a central criticism in both. There was no real support operation behind the product and no feedback coming out of it, and the Better Business Bureau graded the result a D.

Turner Sports and NBA Digital answered by creating the role, new that season, to close a gap in product strategy. I was hired into it in February 2014, midway through the same season the articles were written about.

Deadspin · Tim Marchman · 3 Nov 2013

“Don’t Pay For NBA League Pass Broadband, It’s Straight-Up Garbage”

Read the article

Deadspin · Timothy Burke · 15 Nov 2013

“NBA League Pass Broadband 2013-14: Still Shitty, You Should Not Buy It”

Read the article

2The build

Customer satisfaction, run as a product

A product gets a service model, instrumentation, KPIs and a release cycle. So did this one. I designed the support model end to end: channels, routing rules, agent tooling, the communications platform, and the KPI framework for every fan-facing interaction across 16 platforms, including web and the platform’s first live phone support.

Then I made the contact itself carry data. Oracle Service Cloud was wired by API to nba.com and League Pass, and the tagging requirements I wrote routed each chat by NBA identity and the page the fan was on. A question arrived with its context attached.

3The load

Opening night is the test, every year

Support volume in this business is seasonal and it is sharp. Season openings carry roughly four times the daily contact load of the regular season, and a single outage can beat that. The job was to know that number before the night arrived, not after.

So the operation kept three seasons of daily contact volume, split by preseason, season start and regular season, and planned vendor agent headcount against attrition before each opening night. Incidents were counted per game beside video starts, so a bad broadcast could be named rather than argued about.

Two people at desks with laptops, desk phones and headsets, facing a wall of monitors showing live NBA games, broadcast feeds and product pages.
Game night at NBA Digital. The wall carried live games, broadcast feeds and the product’s own pages. The desks below carried the phones and headsets.

4The instruments

A reporting rhythm the whole organization read

Measurement only changes behaviour when it arrives on a schedule. The operation ran a daily view of yesterday, a weekly view of the past week, a season view cumulative from opening night, and event dashboards when something unusual happened. One classification model sat underneath all of them, so the daily number and the season number were the same number.

The weekly report is the plainest example. It carried contact volume by day, video starts against it, the channel mix, the device mix, the broadcasts that drew the most viewing, and the games that drew the most incidents. Customer KPIs went to NBA executive leadership on that cadence.

A wall monitor showing the live support dashboard: incident sources by hour and channel, a device mix donut, and tables of user issue, agent action and billing action.
The live dashboard on the wall, November 2019. Incident sources by hour and channel, the device mix, and every contact resolved into a user issue, an agent action and a billing action.

5The ordeal

A league that stopped, then restarted

The NBA suspended play in March 2020 and returned in a bubble that summer. League Pass sits inside one of the most visible brands in American culture, so a subscription question that year was never only a subscription question. It was a chaotic stretch to manage, and there is no substitute for having managed one.

It landed at the hardest point in the staffing year, late in the season with the support floor at its thinnest. Then the floor itself had to move. Getting agents home and working meant sorting the logistics of laptops and access before anyone could answer a single fan.

6The model

Prioritization from what fans wrote

The last piece was deciding what to fix first. I defined and led a text analytics study that detects keywords in survey verbatims, scores their sentiment in local context, then compares each keyword against the sentiment of the question overall and ranks the difference, so the words moving the score rise to the top.

It was prototyped on one season, then run in production the next against a much larger body of text. The output fed the product cycle rather than a report: season-start fixes from September to November, roadmap delivery from December to February, feature testing from March to May.

One path from a fan’s question to the roadmap

  1. 01

    Contact

    Chat and the platform’s first live phone line, across 16 platforms, including web.

  2. 02

    Routed

    Oracle Service Cloud, wired by API to nba.com and League Pass, routing on identity and page.

  3. 03

    Classified

    A governed contact reason taxonomy and NLP tagging turning raw sentiment into metrics.

  4. 04

    Clustered

    Feedback clustering models, built with engineering and analytics on that one classification.

  5. 05

    Prioritized

    Fan evidence into roadmap prioritization, and customer KPIs to NBA executive leadership.

7The return

What the operation gave back

The Better Business Bureau rating went from D to A within two years. The season-opening takedowns stopped while I held the role. Preseason contact volume fell by more than a third across three seasons, even as the subscriber base grew and new products launched on the same support model.

The durable part is the measurement. Everything below existed because the role built it, and all of it outlasted any single incident.

The measures

What the reporting could finally answer.

Some of this existed in scattered form before the role, and some of it did not exist at all. What changed is that every question below could be answered from the same classified contact record, on a schedule, by anyone who needed it.

QuestionAnswered by
How much load will opening night carry?Three seasons of daily contact volume by season phase
Is the product getting easier to use as it grows?Contact volume measured against video starts
What is going wrong right now?Trending issues, incidents per game, and per platform
What do fans ask, and what actually resolves it?Contact reason and agent action, on one taxonomy
Why do subscribers leave, and who is at risk?Cancellation reasons and post-incident survey by usage segment
What should product fix first?Keyword impact scored against satisfaction

Scope

This role owned customer satisfaction. Other teams built and ran the streaming platform itself. Everything described here is support, feedback, sentiment, and the relationship with the fans. Subscriber growth is context, not a claim that this operation caused it alone.

Sources

Contact

Want the longer version?

I’m happy to walk through the taxonomy, the routing design, the incident dashboards or the text analytics model in detail.

Get in touch
linkedin.com/in/aaronzeanah Résumé (PDF, 2pp) Back to the portfolio