Twenty-One Years of Weird, Wonderful Music: and What the Data Says

Around eighteen and a half years ago, I started a new job as a business analyst. I started in mid-December and was thrown straight into a full-on project. Because I had no annual leave and plenty of work to get on with, I was back at my desk after the Christmas/New Year break very quickly.

Those first weeks of early January can be a lonely business in an office. This was before we all had Spotify on our phones or could stream music on work devices, so I brought in a little radio and tuned into RNZ National.

What I found, on those quiet summer afternoons, was something magical.

Phil O'Brien started it as a summer fill-in; Simon Morris joined him a few years into the run. Two guys who just played whatever they wanted. And I mean whatever. They played the most amazing and sometimes the most objectively ridiculous songs. They had themes, and played songs with sometimes the most tenuous connection to whatever their theme was. They laughed. They joked. They made fun of songs and each other. They had a genuine and deep respect for the idea that sometimes the best songs are the strange, unusual ones: the ones that fell through the cracks, or that most of the world has forgotten. They egged each other on and tried to one-up each other with what they could find and play.

That dynamic had been years in the making. Before the show existed, they had been visiting each other's homes to stage friendly musical duels, trying to outdo one another with forgotten or obscure gems — building up a catalogue of discoveries, as Phil put it, "a good 10 years before we even thought about doing Matinee Idle." Phil's wife had her own description: "student radio for the elderly. It's just two old guys fighting for control of the CD player really."

I was hooked. And I wasn't alone; this "fill-in" radio show gathered something of a national cult following. I would look forward to the summer break and long weekends, and tune in for the afternoon.

And then this year, after 21 years of glorious madness, they wrapped up. Phil O'Brien's parting words were characteristically pointed: "We've spent 21 years playing this music to people and now artificial intelligence has come along and is making music so we don't know what's real anymore... That's why we're leaving today, because we're saying AI has won. We've done our best."

What they left behind was a gap in my holidays, a gap on the airwaves and, after a bit of digging, a complete archive on the RNZ website listing every single song they played across two decades.

I was intrigued. I wanted to go back and relisten to some of those gems. I wanted to know what the most popular song they ever played was, or their most-featured artist, or their favourite genre or decade. I wanted to relive some of the cover versions they would often end each show with.

So, in the spirit of my other high effort, low value projects, I decided to do something about it.

Step One: Getting the Data

First, I scraped the RNZ archive and pulled out all the songs. And then I tried to make sense of what I had.

With 21 years of shows in the archive, the data was all over the place. Band names had misspellings and inconsistencies, because when a human being types tens of thousands of entries over two decades, things drift. I found 140 distinct inconsistencies in artist names alone. Dr Hook vs Dr. Hook. Devo vs DEVO. Half Man Half Biscuit vs Half Man, Half Biscuit. The Shangri-Las vs The Shangri-las. And my personal favourite: Billy Bragg appearing as "BIlly Bragg" (capital I where it shouldn't be), which must have been an autocorrect that snuck through and persisted for years.

Beyond the naming chaos, the data was sparse. There was no information about when songs were released, what genre they were, how popular the artist was, or really anything beyond the artist name, the song title, and the date it aired.

I spent months trying to fill in those gaps. I used the Spotify API, Last.fm, MusicBrainz, and Discogs to enrich the dataset with release years, genres, and popularity metrics. I made real progress, but a long tail of songs simply couldn't be matched to anything in any database. The remaining gaps I eventually closed using an AI classification pass, working through each unresolved song by artist name and title alone and assigning it to the nearest genre from a canonical list.

The result: 15,411 songs, 5,722 unique artists, and 21 years of one of the best radio programmes New Zealand ever produced, captured in a single dataset.

Here's what the numbers say.

The Most-Played Song: Cousin Mosquito

Here's where the data delivers its first twist, and also a perfect illustration of the data quality problems I dealt with.

The most-played song across the entire archive is Cousin Mosquito by Congresswoman Malinda Jackson Parker. Played 16 times in total.

But here's the thing: I almost missed it. The song appears in the dataset under four different artist name spellings:

• Congresswoman Malinda Jackson Parker
• Congress-Woman Malinda Jackson Parker
• Congresswoman Malinda Jackson-Parker
• Congresswoman Melinda Jackson Parker (Melinda, not Malinda)

And split across two titles: Cousin Mosquito and Cousin Mosquito #1.

A naive count treats each variant as a separate artist. Grouped correctly: 16 plays. The most loved song in the entire archive, and it was hiding in plain sight behind four slightly different spellings of the same name.

The Artists: A Hall of Fame and a Hall of One-Timers

The Beach Boys edging The Beatles feels about right for Matinee Idle. But Sparks and Harry Nilsson tied at third — ahead of the Rolling Stones — is quintessential. And the presence of Half Man Half Biscuit and Burton Cummings in the top ten tells you everything about the show's sensibility.

What's equally striking is the picture underneath that top ten.

Of the 5,722 unique artists in the dataset, 3,620 of them (63%) appeared exactly once. The dropoff after that is steep: 809 appeared twice, 339 three times. Only 76 artists — just over 1% — made it into double figures. The whole ethos of the show in a single chart.

The Time Machine

The show was famous for playing old stuff. Not just "this 90s classic" old; genuinely old. The data bears this out.

The oldest song in the dataset is from 1930: Puttin' on the Ritz by Fred Astaire, which aired 91 years after it was recorded. Close behind: Leola Manning's Satan's Is Busy in Knoxville (also 1930) and a 1931 recording of Dream a Little Dream of Me by Ozzie Nelson, the original, decades before Mama Cass made it famous.

This proves Simon and Phil never favoured one decade over another — they pulled from across the full century of recorded music almost exactly evenly.

The Genres

With 100% genre coverage across all 15,411 songs — achieved through Spotify, Last.fm, MusicBrainz, Discogs, and a final AI classification pass — the full picture looks like this:

Rock, Pop, and Classic Rock lead comfortably. But the presence of Comedy at 734 songs — nearly 5% of the entire archive — is the one that tells you most about the show. Alongside it: Blues (588), Indie (623), Progressive Rock (519), Rockabilly (378). And at the bottom, 55 songs classified as Classical.

How that mix shifted across the show's fifteen years in the archive is where the more interesting story lives:

The Rock and Classic Rock foundation (red and orange) holds steady across the full fifteen years — together accounting for roughly 35% of the playlist in almost every year. Below that stable base, a few clear movements emerge. Pop (yellow) grew its share gradually and by the final years had clearly overtaken Classic Rock for the consistent second spot. Folk (teal) crept up quietly from around 8% in 2009 to over 10% by 2020–21 — the show getting slightly rootsier as it aged. Comedy (pink, top band) peaked at over 9% in 2013 and had shrunk to under 4% by the final two years — either the well of genuinely funny songs ran dry, or the tone shifted as the end approached.

The Songs Databases Don't Know

A small sampling of the weirdest, rarest finds — so strange and unknown that no database could place them. They're in the dataset waiting to be discovered again, but you'll need to do the discovery work yourself.

• Father Ted ft. Father Dougal — My Lovely Horse
• William Shatner & Joe Jackson — Common People
• Coby Smulders & Nicole Scherzinger — Two Beavers Are Better Than One
• Simon Morris — Walking in Guildford
• Simon Morris — The Hobbit

A note on Spotify

The explorer links each song to Spotify. If you have Spotify Premium, clicking play streams the full track. If you don't, you'll get a 30-second preview clip — long enough to know whether you want to go find the full version somewhere else.

I built this with Spotify because I use Spotify and I build these things primarily for myself. If you'd strongly prefer another provider — YouTube, Apple Music, Last.fm — I'm open to it, just reach out and we'll see what we can work out. Send me a message on Mastodon or LinkedIn. Bribes help.

Final Thoughts

The data is clean, the genre is assigned for every single song, and I've built an explorer so you can dig through all of it yourself. Search by artist, year, genre, or decade, or just hit the random button and see what comes up.

A caveat worth making: some songs played over the years will have been missed. Some may never have made it into the RNZ archive. Some may have slipped through when I scraped it. Some were probably casualties of the months of cleaning, merging, and reclassifying I did to get here. The 15,411 songs in this dataset are an excellent sample — but they're not a perfect record.

And some of what made the show genuinely great isn't in here at all. The themes Phil and Simon chose each episode. The jokes they made between songs. The tenuous connections they'd find between one track and the next. The banter. Reading out the hate mail. That's all gone — you had to be there. What did survive is, in my opinion, an excellent distillation of the essence of Matinee Idle: the breadth, the obscurity, the comedy, the quiet commitment to songs the rest of the world had forgotten.

Simon Morris put it simply on the final show: "I don't think there's any way in the world you could call this a job. Matinee Idle has been a joy."

If you were a Matinee Idle listener, I hope this brings back some memories. If you weren't, well, now you have 15,411 songs to work through. Start with Cousin Mosquito.

The Matinee Idle archive explorer is at matineeidle.radomski.co.nz