Skip to content

LiveOps Tooling -- Feedback & Improvement Opportunities

PLAYBOOK | DERIVED | Updated 2026-04-08 | Owner: Jordan Wagner (LiveOps Lead)

This is an operational reference for tooling gaps, proposed fixes, and prioritization across NBA Top Shot and NFL ALL DAY LiveOps workflows. The core finding: the limitation is not a lack of tooling but that existing tools require manual, one-by-one operations at a scale where that approach breaks.

Priority 1: Bulk Operations in Retool

The Problem

Most LiveOps workflows rely on Retool. The necessary functionality exists, but most workflows require manual creation or updating of records one at a time.

Creating 128 pack listings/distributions for a single NFL pack drop: preparing the actual distribution data took ~30 minutes. Creating and minting the distributions inside Retool took ~4 hours due to manual entry.

The majority of time is spent:

  • Copy/pasting data field-by-field
  • Creating each distribution individually
  • Clicking "Mint" one-by-one for every distribution

This pattern applies across: Leaderboards, Challenges, Fast Break, Pack Ops (both NBA and NFL).

Proposed Fix: CSV-Based Bulk Creation & Updates

  1. Export current records from Retool to a spreadsheet
  2. Edit or create new rows in the spreadsheet
  3. Upload the CSV back into Retool
  4. Retool processes the records and applies: creations, updates, field changes based on record UUIDs

Expected impact: Tasks that currently take 4+ hours reduced to ~1 hour or less. Reduces manual error, improves consistency, allows LiveOps to prepare operations faster.

Priority 2: Bulk Minting

The Problem

Minting distributions requires clicking "Mint" individually for each record. For events with dozens or hundreds of distributions, this is the single largest time sink in drop execution.

Proposed Fix: Multi-Select Bulk Minting

Allow multiple distributions to be selected and minted in a single operation. Required safeguards:

  • User-controlled mint order -- operator defines the sequence
  • Sequential processing -- respects the defined order
  • Failure safeguards -- if a mint fails, the process stops (prevents downstream cascading)
  • Resume functionality -- ability to restart minting from the failed distribution

Priority 3: Render Pipeline Observability

The Problem

When render jobs are triggered for parallels or subeditions, there is no centralized visibility into job status after initiation (from a Retool perspective).

From the NFLAD 2025 Season Retrospective:

"Failure rate on NFL during this period, depending on the batch run, was 30-50% higher than what we typically experience, causing major headache."

"To get drops over the line, we stepped in to run a number of these manually on local machines, there were 3 drops in particular with heavy assist as well as spot jobs here and there."

"We ran into a 'domino' failure issue where a certain kind of failure would stall and kill the next render in line, which caused high fail rates in large volume batches."

Root causes identified:

  • Tooling produced unhelpful logs. Improving logging is what eventually helped track down issues.
  • Manual input / naming caused OS-level failures. Long-path and long-set issues hit OS limitations for character and path limits.
  • Domino failures. One failure stalls and kills the next render in the queue. Dominoed renders had nothing wrong with them and passed individually, making the root cause difficult to track.

Without centralized tracking:

  • Difficulty identifying which specific subeditions or parallels failed to render
  • Time spent manually verifying render outputs
  • Risk of missing failed renders before content goes live
  • Inefficient re-render workflows when only a small subset needs retrying

Proposed Fix: Render Job Tracking Dashboard

A tracking interface within Retool (or dedicated internal view) providing:

Render Job History:

Field Description
Job ID Unique identifier
Series / Set Content grouping
Parallel or Subedition Specific variant
Initiated by Operator name
Timestamp When triggered
Job status Queued, Processing, Completed, Failed

Detailed Render Status (per job):

  • Moment / play
  • Subedition / parallel name
  • Individual render status
  • Failure reason (if applicable)

Failed Render Identification: Clear indicators highlighting failed renders without requiring full audit of the set.

Re-Render Functionality:

  • Select individual failed renders
  • Bulk select multiple failed items
  • Trigger re-render directly from the interface

Priority 4: Parallel Asset Management

The Problem (Shared NFL + NBA)

Parallel assets involve managing a large number of files and rendering variations. Current workflows create friction in both asset management and render mutation generation.

From the NFLAD retrospective:

"864 parallel editions" produced across the 2025 season.

"Production of parallels -- very manual process. Should apply the automated process that's being built for NBA."

Proposed Fix: Improved Asset Upload & Organization

  • Automated file sorting based on metadata: set name, tier, player name, parallel/subedition name
  • Consistent templated naming conventions for parallel assets
  • Auto-attachment: With standardized naming, uploaded assets automatically attach to their corresponding plays within the curation tool

NBA-Specific Issue

When generating render mutations, the tool forces all subeditions to render. As a result, Sam must manually edit the JSON, remove unwanted subeditions, and requeue the job in Retool.

Fix: Allow individual subedition selection directly within Retool when generating render mutations. Removes the need for manual JSON editing.

NFL-Specific Issues

  • Special characters in player names cause validation failures in the bulk upload tool
  • Upload tool times out when uploading too many assets at once (parallel sets are large by nature)
  • No automated player-to-parallel-asset linking despite file name validation existing

Structural Improvement: Sets Table for Parallel Asset Configuration

Introduce a Sets Table, particularly for NFL, allowing configuration of series, set, and parallel relationships. This matters because:

  • NBA creates a new set each series
  • NFL reuses sets across series
  • Configurable set relationships would make parallel asset management more scalable

NFLAD 2025 Season: Production Scale Reference

The 2025 NFL ALL DAY season shipped 24 distinct campaigns across ~7 months. Production context for tooling investment decisions.

Season Drop Volume

Archetype Drops Examples
Positional Parallel Sets 6 Wideout Wonders (x2), Launch Codes (x2), Ground Greatness (x2)
Premium Legendary 3 Holo Icon Drops 1-3
Tentpole Events 3 Feast Packs, Banner Year, Rookie Revelation & Sendoff
Postseason 8 Wild Card through Rewind (including Super Bowl LX)
Other 4 Rookie Debut, Fresh Threads, Regal Rookies, Legends of the Turf

Per-Drop Asset Requirements

Each drop typically required a full creative package:

  • Set art and pack art (multiple tiers)
  • Player renders across edition tiers (Common, Rare, Legendary, Ultimate)
  • Parallel tier renders (Ruby, Emerald, Sapphire; postseason added Diamond, Topaz, Obsidian)
  • Email creative (header, body, CTA)
  • Social assets (OG image, announcement, countdown)
  • Blog header and background art
  • Pack breakdown / pricing graphics
  • Trade-in banners
  • Challenge reward graphics

Recurring deliverables on top of drops: Playbook weekly visual suite, weekly Rookie Ladders, drop calendars, social templates, partnership creative (9 partnership campaigns documented).

Key Production Findings

Parallel artwork: ~6 hours/week. Each parallel requires 3 separate assets that cannot be fully automated.

Sizzle reels: ~6 hours/week, ~24 hours/month. Used on social media and in-app. No engagement data comparing sizzle reels to static assets.

NFLPA approval: Always 2 business days. External dependency that compresses production windows.

Render failure rate: 30-50% higher than typical during the NFL season. Three drops required heavy manual intervention including running renders on local machines.

Content Art Direction Feedback

From the retrospective discussion:

"Overall good art direction. Standouts: RD, rookie marquee, banner year. Could use some art styles from NBA Top Shot. Less on words, more art style. Put energy into the higher end set arts."

On pack art:

"Labels/stickers helpful. Maybe collectors could see parallels label. 6 players to 3 players was a big win because there is more space for players."

On parallel tiers:

"Topaz and diamond were the favorites by users. Ruby/sapphire/emerald -- they didn't feel as special. Regular season to post season was a progression look."

On autographs:

"Signatures look good. Use NBA approach. Signature editions."

Season Financial Markers

  • Super Bowl week: Largest revenue event -- $186K, 2,613 buyers
  • Parallel burns: 39,904 Rare burns and 852 Legendary burns -- highest ever for NFLAD
  • Holo Icon market impact: Post-distribution Legendary prices dropped 53-56% (potential over-supply signal)

Summary: Three Investment Areas

The tooling feedback consolidates into three categories, in order of operational impact:

1. Bulk Operations

  • Bulk record creation/updates via CSV workflows
  • Bulk minting with sequential processing and failure safeguards
  • Expected impact: 4+ hour tasks reduced to ~1 hour

2. Asset Management

  • Automated file sorting and attachment by metadata
  • Larger batch upload capacity
  • Standardized naming conventions with auto-linking
  • NFL Sets Table for configurable parallel relationships

3. Rendering Workflow

  • Subedition selection in Retool (remove forced all-subedition renders)
  • Render job tracking dashboard with failure identification
  • One-click re-render for failed items
  • Fix domino failure pattern in batch rendering