Learn how we approach hold music
They were made for human voice. The traditional voice band on telephones runs in the range from 300Hz to 3400Hz. Try to keep your music from going above or below these ranges, because it will typically manifest itself as silence. In addition, telephone networks were designed for 8kHz sample rates which (per some fancy math called the Nyquist-Shannon sampling theorem) means the maximum effective rate of 4000Hz, or 4kHz. The 8kHz sample rates also brings additional problems, since most audio is recorded at much higher sample rates and converting it down will quickly lose quality in playback. However, you SHOULD understand how your music sounds when converted down to 8kHz before calling it passable. We typically re-sample a standard file down to 8000hz, speed it up by about 400%, then remove frequency bands by inversion and panning to center (i.e., vocal removal in Audacity.)
When Wideband audio is universally supported, the hold music quality problem we have today will be fundamentally improved.
Unless you're working in a lab environment, phone calls can and will pass through multiple telephone carriers, different PBX systems, across multiple geographic regions, and finally through different telephones and speaker technology. Carriers want to save bandwidth and will down convert your high-quality audio, PBX systems may do the same (often up and down converting codecs on the fly), and every caller's telephone will use hardware playback and/or software optimization for the "optimal" experience. We look at this entire process almost as a Dependency Hell for telephony developers.
Except in some rare circumstances (like a "rock music support line") it will not be possible to ensure that every caller enjoys the genre of music you play. However, when it comes to complaints about hold music, the genre of the music is generally not in question. Your callers use your hold music to gauge status of the call: "Am I still on hold, or is something else happening?" Ever been on hold and heard a silence -- expecting someone to jump on the line or for you to be transferred somewhere? Consistent, rhythmic hold music helps instil confidence for your callers as they wait.
We use SoX and Audacity for our optimization efforts. Audacity when we have to convert a song by hand to get it to sound right, and SoX for most batch operations and the initial "test version" of new sound files. Do you use or recommend other tools in your quest for the perfect sound file? Let us know which ones!
There are many artists out there creating wonderful content and releasing it under Creative Commons Attribution licenses. This simply means: use it as you please, just give me credit and pass along this license. In the real world, these gifts to the community get used without attribution all too often. If you use someone else's content and they simply ask that you attribute them, then please do so! Holdmusic.io originally started as an easy way to provide attribution for CC licensed hold music in our own phone systems (via an iframe or direct link in our web applications and website.) We provide the same functionality in our premium account and still encourage those using only public domain music and recordings to include attribution.