Nugs.net changes how fans discover live music with custom recommendations personalized by AI.
Client profile
A leading live music streaming platform in the US
Industry
Other
Region
North America
Improvement in user retention
Increase in paid plan conversions
Nugs.net hosts the world’s largest catalog of live concert recordings. Over 250,000 tracks across 15,000 shows. Metallica, Pearl Jam, Dead & Company, Billy Strings. Rock, jazz, bluegrass, country. The platform started in 1993 as a tape-trading hub for Grateful Dead fans. Three decades later, it serves a global listener community chasing the real thing. Not studio polish – the energy of a specific night in a specific venue. The catch: new content only arrives when artists tour.
01 The ChallengeIn music streaming, about 18% of users switch platforms every year. Monthly churn rates have climbed from 2% to over 5% since 2019. Spotify and Apple Music stay fresh year-round with new albums and editorial playlists. Nugs.net does not have that option. Its catalog grows when artists play live. When touring slows down – off-season, between album cycles – the content pipeline slows with it.
That seasonality hit every part of the business. A product manager tracking weekly active users could see the pattern in the data. Engagement dipped in winter months, recovered during summer festival season, dipped again. A growth marketer watching trial-to-paid conversion rates saw the same curve. Fewer new shows meant fewer reasons for trial users to subscribe.
Provectus, an AI-first systems integrator, has partnered with Nugs.net for over eight years. The team saw an opening in the behavioral data. The numbers told a clear story. Fans who listened to more than four artists per month retained at far higher rates. Users who explored older recordings – archived shows from past tours – stayed on the platform longer. Trial users who followed multiple artists during their free period converted to paid plans more often.
The catalog held decades of live recordings. The bottleneck was discovery. Most users stuck to the two or three artists they already knew. They never found the 2003 bluegrass set or the jazz trio’s live album from last summer. The content was there. The path to it was not.
The team set four goals. Improve content discoverability. Increase listening diversity. Give the product team visibility into catalog performance. Grow the number of a mrtists each user follows.
02 The ApproachOff-the-shelf recommendation tools couldn’t handle Nugs.net’s data. Standard models struggle with the signals that matter here. A single artist might have hundreds of live recordings, each slightly different. User behavior skews toward deep, repeated listening rather than casual browsing. The signals that indicate taste look different here. A fan replaying a specific encore. A listener skipping a studio cut to find the live version. Standard models miss these patterns.
Provectus brought eight years of context on the platform’s data, its users, and its product architecture. The team started with workshops to map user flows, define business hypotheses, and set measurable KPIs. Then they moved into experimentation.
The core question was how to balance accuracy against diversity. A recommendation engine that only suggests similar artists keeps users comfortable but does not expand their listening. One that pushes too far afield loses trust. Provectus ran extensive A/B tests, measuring each variant against retention and conversion targets.
A key finding shaped the final design. Different users had different appetites for discovery. A Dead & Company fan might welcome a Billy Strings suggestion. A Metallica listener might not. The engine needed to adapt its range to each user’s behavior. A single model applied to everyone would not work.
Provectus delivered the MVP in three months on a lean budget. Fast enough to test the primary hypotheses and adjust the roadmap before committing to a full rollout.
03 The BuildThe recommendation system adapts its output per user. It reads listening history, follow lists, replay patterns, and skip behavior. Then it calibrates how far to stretch each suggestion. A user who already listens to six genres gets broader picks. A user who replays the same three shows gets gentler nudges toward adjacent artists.
The AI team at Provectus worked alongside product designers, product managers, and user researchers throughout the build. Recommendations had to land well in the UI, not only in offline metrics. Where a suggestion appears matters. When it appears matters. How it is framed affects whether a user clicks or scrolls past.
Quality control tied AI performance directly to product KPIs. The team built a three-part evaluation loop. System-level metrics covered precision, recall, and diversity scores. Product-level metrics tracked click-through, listen-through, and follow rates. The Nugs.net editorial team handled expert review. If the model scored well technically but users did not respond, the model got tuned. If users responded but the editorial team flagged poor genre coverage, it got tuned again.
Behavioral analytics drove every product decision. A typical test: do home-feed recommendations generate more engagement than post-playback ones? Each hypothesis was tested and either confirmed or dropped. The projected ROI for each initiative was calculated before development started.
04 The ResultsThe numbers arrived fast. Users who engaged with AI recommendations retained at 40% higher rates than those who did not. Trial-to-paid conversion doubled.
40%
Higher retention
2x paid conversions for users engaging with recommendations
Listening diversity broke through the four-artist-per-month ceiling. Users began playing up to 15% more archived shows from past tours and seasons. Artist followings nearly doubled, then stabilized at the new rate. Daily retention climbed close to 2x its previous baseline.
The projected ROI for the MVP hit 183% within twelve months of release. That number reflected coordinated work across product, design, mobile, web, backend, and AI teams at Provectus.
For the business, the most important shift was structural. Nugs.net’s growth no longer tracks touring schedules alone. The recommendation engine surfaces older recordings and cross-genre connections that keep fans engaged between tours. Content that sat dormant in the archive now generates daily listens.
05 What’s NextNugs.net can now extend personalization into search, notifications, and artist pages. Event-based features that connect streaming to live shows are next. The eight-year partnership with Provectus gives both teams shared context to move fast on each new initiative.