Mentor.FM, its shutdown and the pioneering era of Music Recommender Systems

Eugenio Tacchini
4 min readAug 1, 2023

--

Mentor.FM belongs to the first, pioneering, era of Music Recommender Systems. I developed the first prototype in 2011 as an academic experiment (part of my PhD thesis exploring Serendipity in Recommender Systems) and then launched globally in 2013 in collaboration with Deezer, leveraging their extensive music library for streaming.

Mentor.FM was a one-click, frictionless, recommendation radio: you logged in using your Facebook account and in a few seconds you started an immersive musical journey. Mentor.fm built a draft of your music profile according to your FB profile and then continuously learned from your like, dislike, skip signals.

You got a personalized radio channel that combined your favourite artists and other artists and created a flow (akin to an infinite playlist) in which the order of the songs mattered. A sequence more similar to a DJ-curated playlist than a random selection of songs you probably like.

A “Surprise me” mode was also available and aimed at introducing unknown “Musical Worlds” to the users, fostering musical exploration.

Mentor.FM Launch Video (2013)

Looking back, I’ve started a nostalgic journey retracing the history of Mentor.FM and the various features that were added over the years. I think it’s worth mentioning:

  • The iconic synthesized voice of a DJ announcing the song :)
  • MentorMood, a mood-based radio playing music according to your taste and your current mood
  • Concerts recommendations
  • Song dedications
  • Gigromantics, a Mentor.FM-based app aimed at connecting people sharing the same musical taste

I have seen some of them implemented, eventually, in other major streaming services.

During the Mentor.FM era, I was also active in the Music Hackathon community, building or contributing to build several applications, some of them connected to Mentor.FM.

Here is a partial list:

“/her music”: a conversational, voice-controlled, browser-based Juke box inspired by the movie Her

“The World Beat” a mood-of-the-world thermometer based on what people were listening to

“The Daily Beat”, a personalized music newspaper app

The Music Hackathon scene was amazing: you typically spent the weekend in a vibrant European capital, immersed in a creative atmosphere, coding and hanging out with fellow music hackers who shared the same passion for innovation and music. One of the most important appointment was that in Barcelona, held during the SONAR music festival.

Business model

I have never found a proper business model to support Mentor and honestly I have never seriously looked for external funding: passion for what I was building was my primary driver.

As time passed, however, I started to dedicate less and less time to Mentor and more and more to another project of mine, DaDaBIK, a low-code no-code platform with a solid business model and a long history of development. As a result, my focus has shifted towards growing DaDaBIK, a project that has thrived on both passion and a viable foundation for success.

I joined the Deezer affiliation program, which was supposed to provide a fee for each new Deezer subscriber I referred. Unfortunately, after the initial payment, they disappeared, the made the affiliation data opaque and very difficult to read. Moreover, they ceased responding to any affiliation-related emails, leaving me with a rather unpleasant experience.

Recent period and shutdown

At some point I had to do some work to update Mentor to the latest technologies, I started working on it but then I postponed it several times and then stopped because of lack of time and motivation.

You have probably seen the “Mentor.FM is under maintenance” message in homepage since two years ago :(

So, today Mentor.FM is officially suspended indefinitely. I don’t know if, in future, it will reemerge in a different form or powered by another streaming service.

As software developer, working on the Mentor.FM recommendation algorithm, trying to make it smarter, has been the most engaging, stimulating, rewarding and fun working experience, combining my passions for Computer Science and Music.

The period between 2010 and 2011, when I started working seriously on these topics, was a truly exciting time. The field of recommendation algorithms for music was full with unexplored aspects, making it feel like a pioneer’s playground.

During that period, two very important players were Last.FM (an amazing service!) and Pandora. Nowadays recommendation algorithms are integrated in music streaming services, but during their first era they often operated independently, solely dedicated to perfecting the art of music recommendations.

Replacements? Ideas for a renewed Mentor.FM?

I am the author of Mentor.FM, but I was also an avid user of the platform, I used it daily as my personal DJ and I have discovered tons of music thanks to it so I miss it also as a user, both the entertaining side and joy of musical discovery it provided.

I have never found a suitable replacement. The most similar thing at the moment is probably Deezer Flow, but honestly I think the playlists generated are not as interesting and entertaining as the ones provided by Mentor.FM’s curation.

If you have any suggestions, ideas regarding a possible future for Mentor.FM or simply want to share your thoughts or memories about Mentor.FM, please leave a comment here or send an email to eugenio@dadabik.com

Mentor.FM, when active, had a small yet devoted and passionate user base. I think it was a niche, addictive, little product. Over the years, I have received countless messages, ideas, suggestions, and comments from the community, many of which have been incredibly valuable.

Dear users, sorry for the lack of communication during this last period and thanks to all of your for the support and enthusiasm, Mentor.FM has been an amazing experience for music enthusiasts as we are.

--

--

Eugenio Tacchini

Passionate about Computer Science, Music, Cinema. Lecturer of Databases & Data Science. Founder of https://dadabik.com (one of the first low code platforms)