A
OVERHAUL
A new schedule announced, all at once
- maximum visibility of the change
- simultaneous loss of L1 markers
- L2 gain not consolidated
- not published
CREATIVE INDUSTRIES · TRANSFORMATION
“How can a music radio station change its programming every week without losing the core listeners who keep it alive, when every change is a bet you only measure three months later?”
A French-language music radio station, established in its market for more than twenty years, had plateaued at a 4% audience share. A loyal but aging core of listeners, a slightly declining cumulative audience, stable time spent listening. The programming director had a periodic music test and a format study every few years. Between two studies, the station was run on instinct.
The question put to the system was not about a schedule. It was about how to evolve a station continuously, without waiting for the next ratings wave to find out whether a choice was wrong.
- AUDIENCE SHARE
- 4% → 7.5%
- OBSERVATION HORIZON
- six months
- SYNTHETIC AUDIENCES
- two, L1 and L2
- CADENCE
- weekly
THE PROBLEM
A radio station cannot
test its programming
on its listeners.
Every failed attempt
is a reason
to switch.
The programming director knew what she wanted: refresh the music, lighten the morning show, give more room to live content. She also knew what a badly received change costs. In radio, listeners do not complain; they turn the dial, and the ratings only say so three months later.
The music test ranked songs by familiarity and burn. It said nothing about what a song does in the middle of an hour on air, nor about the difference between the listener who stays and the one who leaves. Previous schedule overhauls had left a precise memory: an audience dip before any gain, and the temptation to go back before the gain arrived.
WHAT THE DECISION HAD TO RESOLVE
- THE CORE AUDIENCEhow many loyal listeners are lost with each change on air?
- CONQUESTwho are the listeners who pass through without staying, and why do they leave?
- THE DELAYhow do you know, before the next wave, that a programming choice is working?
- THE MUSICis a song that tests well a song that keeps people listening?
- THE IMAGINGjingles, sweepers, promos: where is the saturation point?
- LIVE CONTENTdoes live content bring audience, or only image?
A MECHANISM: AUDIENCE SHARE IS A PRODUCT
A number of listeners
multiplied by
a time spent listening.
The two are not won
with the same listeners.
Cumulative audience depends on those who come. Time spent listening depends on those who stay. A station at 4% aiming for 7.5% does not double its number of listeners; it gains some, and keeps each of them listening longer. Two populations, two behaviors, and almost two stations.
A DISTINCTION: L1, L2
Two ways
of listening to
the same station.
The system built two distinct synthetic audiences and questioned them separately, every week, on the same on-air elements.
- L1listeners for whom the station is their first choice: the one switched on by default, the one that carries most of their listening time. Loyal, older than the market average, attached to the station’s historic sound, sensitive to any loss of familiar markers.
- L2listeners for whom the station is a secondary choice: heard in passing, alternating with others. Younger, quick to switch, exposed to competing stations and to streaming, sensitive to repetition and to talk blocks. Most of the growth potential.
- THE GAPquestioned on the same elements, the two audiences rarely gave the same answer. That is precisely what the station had to learn to manage.
WHAT WAS TESTED
An entire station,
element by element,
week after week.
The case is not a single test but a permanent loop. Every week, the on-air elements planned for the following week are submitted to both audiences before broadcast. Every month, the audiences are asked not what they think of the station, but what they would like to hear on it.
- 01SONGS, BEFORE ENTERING ROTATIONeach candidate song is placed within a simulated hour of programming, between the songs before and after it, rather than tested in isolation.
- 02STATION IMAGINGjingles, sweepers and promos, tested for density (number per hour) as much as for content.
- 03SHOW FORMATSmorning show, daytime talk, evening configurations: each version submitted to both audiences before going on air.
- 04LIVE CONTENTsessions, concerts and live broadcasts, tested on two separate dimensions: the audience of the slot, and the image of the station.
- 05WISHESevery month, an open question to both audiences about what is missing on air, fed back into the following period’s tests.
Both audiences were questioned individually, in dialogue, on each week’s elements. Results were available before the weekly programming meeting. Delivery never waited for a ratings wave.
THE POPULATION
A listener
is not
an age bracket
with a music genre.
The two populations were built for the station, calibrated on the market’s published audience data, on the station’s proprietary usage data (app, streaming, podcasts), on its previous music tests, and on public socio-demographic data. No personal data entered the system.
They are structured by relationship to radio, combining age, listening moment and location, number of stations listened to, exposure to streaming, tolerance for talk, and appetite for musical discovery.
STATION RANK
first choice or fallback station: the same programming is not heard the same way
LISTENING MOMENT AND LOCATION
car, kitchen, office, worksite: the switching threshold changes with the place
NUMBER OF STATIONS LISTENED TO
one, three or five: the multi-station listener is always comparing
EXPOSURE TO STREAMING
the personal playlist sets the expectation on repetition
TOLERANCE FOR TALK
from the listener who tunes in for the voice to the one only waiting for the next song
APPETITE FOR DISCOVERY
recognizing or discovering: two pleasures, two listening durations
These configurations make part of the population’s heterogeneity visible. The simulation involves synthetic individuals, not a handful of persona types.
REACTIONS
The core audience’s favorite song
is the one that makes
the others leave.
- 01
L1 AND L2 DO NOT SWITCH FOR THE SAME REASONS
L1 leaves when a marker disappears: a historic jingle, a host, a moved appointment. L2 leaves when something lasts too long: a song repeated too often in a day, a talk block that runs on, a promo sequence. For L2, the first minutes of listening decide the rest.
- 02
A SONG’S SCORE DOES NOT SAY WHETHER IT KEEPS PEOPLE LISTENING
Songs rated very highly by L1 in isolated tests, once placed in heavy rotation, drive L2 away. The reverse exists: moderately rated songs that never burn hold listening longer. Rotation was rebuilt on retention, not on stated preference.
- 03
IMAGING SATURATES BEFORE MUSIC DOES
Beyond a certain number of imaging elements per hour, L2 listening time falls, whatever the content of the jingles. Short, regular jingles, better accepted by L2, leave L1 indifferent on one condition: that the historic sonic signature is kept.
- 04
LIVE CONTENT DOES NOT BUILD AUDIENCE, IT BUILDS L1
Live content draws a smaller slot audience than the music programming it replaces. But among the L2 listeners exposed to it, it raises the likelihood of making the station their first choice. Live is an instrument of conversion, not of instant audience.
A MECHANISM: CONVERT, DON’T RECRUIT
Audience share is won
by changing the station’s rank
among those who
already listen.
An L2 listener who becomes L1 does not count once more in cumulative audience. They count several times more in time spent listening. That shift in rank is what the station set out to produce, week after week.
- LEVEL 1 · “IT’S ONE OF MY CAR PRESETS.”one station among others, heard occasionally
- LEVEL 2 · “I STAY TO THE END OF THE MORNING SHOW.”listening time lengthens, switching becomes rarer
- LEVEL 3 · “IT’S MY STATION.”the station becomes the first choice, and its share of listening time with it
A MECHANISM: CO-CREATION
An audience asked
what it wants proposes
what no test
would ask.
The monthly open question fed part of the formats tested over the period. Three of them made it on air.
- A LATE-EVENING SLOT WITH NO TALKrequested by L2
- A WEEKLY APPOINTMENT OF SONGS “YOU NEVER HEAR ANYMORE”requested by L1, placed outside peak listening hours so as not to weigh on L2
- THE SATURDAY LIVE SESSION REPLAYED IN A SHORT VERSION DURING THE WEEKproposed separately by both audiences
COMPARISON
The options,
read on four dimensions.
OPTIONL1L2TIME SPENT LISTENINGAUDIENCE SHARE AT SIX MONTHS
01Programming kept as is, historic schedule and rotation markers preservedmarkers preservedcontinuous erosionstablenot published
02Full schedule overhaul in one go, the option initially consideredsimultaneous loss of markersgain not consolidatednot publishednot published
03Continuous evolution steered by the two audiencesmarkers keptL2 → L1 conversionnot published7.5% observed
Only options for which the case documents measurements carry figures. The mention “not published” is kept rather than an estimate.
EVOLVE
every week, one element changes, tested before broadcast L1 markers remain, L2 reasons to leave disappear one by one
OVERHAUL
a new schedule, one Monday in September L1 lose their markers at the very moment L2 discover a station they do not recognize either
THE CALL
What the decision
kept.
- TO PRESERVE
- The historic sonic signature and the appointments that give L1 their markers.
- TO CHANGE
- Music rotation, rebuilt on retention rather than on stated preference.
- TO REDUCE
- Imaging, brought below L2’s saturation point.
- TO ADD
- Live content as an instrument of conversion, placed outside peak listening hours.
- TO HOLD
- The weekly cadence, with no return to one-off schedule decisions.
OBSERVATION
Six months
after launch.
Management chose continuous evolution. No schedule overhaul was announced. The station changed every week, element by element, each change having been submitted to both audiences the week before. The loop is still running.
This value describes what was observed six months after the decision. It establishes no exclusive causal link between the chosen method and this movement. The case will be updated with subsequent ratings waves; no projection beyond that is published.
- AUDIENCE SHARE AT SIX MONTHS
- 4% → 7.5%
- CUMULATIVE AUDIENCE
- not published
- AVERAGE TIME SPENT LISTENING
- not published
- LISTENERS NAMING THE STATION AS THEIR FIRST CHOICE
- not published
LESSON
The problem
was not
changing the station.
It was never changing it
all at once.
Loyal listeners did not refuse change. They refused to lose all their markers at the same time. Occasional listeners were not asking for a different station; they were asking to stop being given reasons to leave. A station tested every week on two distinct audiences evolved further in six months than an overhaul would have, without the dip that overhauls bring.
The second lesson is about method. The value of permanent testing lies not in a result but in the cadence. A test that arrives before the programming meeting becomes part of how the station is made. A study that arrives every two years stays on the side.
POSSIBLE FUTURES
One station.
Three ways to make it evolve.
B
KEEP
Historic programming, adjustments by instinct
- L1 preserved
- continuous L2 erosion
- stable listening time
- not published
C
EVOLVE CONTINUOUSLY
Two synthetic audiences, weekly cadence, monthly co-creation
- L1 markers kept
- L2 → L1 conversion
- 4% → 7.5% observed at six months
