Skip to content

guide

Can a radio station know its actual listenership?

Actual radio listenership can only be estimated, because FM leaves no record of who received it. A fair survey gives a figure with a margin of error, and that range is what a station should quote.

Takwimu editorial team, Editorial team8 min read

Actual radio listenership can't be counted exactly, because FM leaves no record of who received it, so a station can only estimate it from a fair sample of people and state a margin of error. That estimate can be very good. A station that quotes it as a range, with a period and a population, has a stronger claim than one quoting a large round number.

Key points:

  • FM is one-way, so no system at the station can see a listener.
  • Coverage, streams, replies and advert logs are exact counts of other things.
  • A survey of 1,000 people gives a reach estimate with a margin of about plus or minus 3 points.
  • The period matters: national radio use was 49% daily and 75% weekly in April to June 2026.
  • Quote a range, a period and a population, and never add sources together.

Why can't a station count its listeners?

A transmitter sends a signal and nothing comes back. Ten people or ten million could be tuned in, and the equipment behaves the same.

A TV set-top box or a streaming app can report back to a server. A radio can't. Every figure for FM listening therefore comes from observing some people and reasoning about the rest.

What does a station know for certain?

Five things. None of them is listenership.

Source What it counts exactly Why it isn't listenership
Transmitter and signal data Who could receive you Being in range isn't listening
Streaming server Connections and hours Online only, and it sees devices
Phone lines and SMS Replies Volunteers, with no known rate
Advert log Spots aired It records the transmission alone
Social media Followers and views Accounts, largely outside broadcast

Each is worth having under its own name. Radio streaming analytics and why on-air responses do not measure reach explain two of them.

What is the closest a station can get?

A fair sample of people. There are two routes: ask them, or meter them.

Meters are the more detailed route, and they carry their own limits. RAJAR says of its phone panel that these techniques "are not perfect", since people may not keep their phones with them, though they generally give a good account of people's actual listening.

For a Kenyan station, the practical route is a recall survey. The regular national services with published methods work this way, as how radio listenership is measured explains.

How do you estimate listeners from a survey?

Five steps, shown here with an invented county.

  1. Define the population: 500,000 people aged 15 and over in your county.
  2. Draw a fair sample: 1,000 interviews spread by sub-county, age and sex.
  3. Ask a clear question: did you listen to this station in the past seven days?
  4. Weight the answers to census totals.
  5. Scale up: if 23% say yes, that is 0.23 × 500,000 = 115,000 people.

So the estimate is 115,000. The next step is saying how sure you are.

What is the margin of error?

It is the price of not talking to everyone, as AAPOR puts it. At a 95% confidence level, it gives the range the true figure is likely to fall in.

The formula for a simple random sample is 1.96 × √(p × (1 − p) ÷ n). With p = 0.23 and n = 1,000, that is 1.96 × 0.0133 = 0.026, or plus or minus 2.6 points.

So the fair statement for this county is between 20.4% and 25.6%, or roughly 102,000 to 128,000 people. The range is the finding. Weighting usually widens it a little, and AAPOR notes that the margin applies only to probability-based samples.

What makes a sample fair?

Each person in the population needs a real chance of selection. That is the line between a survey and a list of volunteers.

In practice, you place interviews across the county in proportion to where people live, and you pick respondents by a rule that doesn't depend on who is keen. You then weight the result to census totals for age, sex and place.

Never recruit respondents by announcing the survey on air. Those who answer are your keenest fans. Your reach will look far bigger than it is.

How many interviews do you need?

Fewer than most people think, up to a point. Margins at the widest case, 50%, look like this.

Interviews Margin of error
400 ±4.9 points
1,000 ±3.1 points
2,500 ±2.0 points
10,000 ±1.0 point

AAPOR makes the same point: past about 1,000, additional sampling error falls only slightly. Quadrupling the sample halves the margin. Sample size for an audience survey works through the choices.

Why do smaller stations need more care?

Because small percentages rest on few people. A station with 5% reach in that sample has 50 listeners among its 1,000 interviews.

The margin is about plus or minus 1.4 points, so its 25,000 listeners could be anywhere from about 18,000 to 32,000. Cut those 50 people by daypart and they stop meaning much. A small station should report weekly reach with confidence, and read daypart results as a direction.

Can coverage tell you listenership?

Coverage only sets a ceiling. Your transmitter may reach 500,000 people aged 15 and over, and nobody is required to tune in.

In the invented county above, 77% of the people in range didn't listen in the past week. That gap is the most useful figure a station owns, because it is the size of the opportunity.

Why can't a national survey settle it for you?

A national tracker has too few people in your area. Ipsos described a Kenyan tracker of 2,500 people in 2022, which would be about 53 a county if spread evenly.

Audiences also vary sharply by place. The regulator's August 2026 report shows radio at 83% in Western and 45% in North Eastern. A county station needs figures weighted to its own county, using KNBS census totals.

Which period should you quote?

State it every time. "Listeners" means nothing without a period.

Nationally, the regulator found 49% listening daily and 75% in the past seven days in April to June 2026. An invented station might have a daily reach of 9% and a weekly reach of 23%. Both are its actual listenership, for different periods, and daily and weekly reach explains why each matters.

What is the difference between listeners and listening?

Listeners are people. Listening is time. A station can gain one and lose the other.

Reach counts each person once, and share counts hours. If a station's reach holds at 23% while its share falls, the same people are staying for less time, which points at programming. Share of audience covers the sum.

Why shouldn't a station add its numbers together?

The same people are in every pile. Here is an invented sales deck: 115,000 survey reach, 12,000 stream cume, 40,000 followers and 3,000 SMS senders, presented as an audience of 170,000.

Most of those stream users and SMS senders are already inside the 115,000. Followers may not even live in the county. The sum is bigger, and it is a worse number, because an informed buyer will now doubt the 115,000 too.

How should a station state its listenership?

In one sentence with five parts: estimate, population, period, margin and source.

Here is an invented example. "Station X reached an estimated 115,000 people aged 15 and over in Y County in the past seven days, 23% of the population, plus or minus 2.6 points, from 1,000 interviews in October 2026."

An advertiser can check that sentence. How a station proves its reach and a reach note for an advertiser show how to take it further.

What if a client asks for the exact number?

Tell them the truth. There isn't one, for any station, in any country, so show the range and how it was reached.

Many stations fear that a range sounds weak. In practice, it does the opposite with a serious buyer, because it shows you know what you measured. The stations that struggle are those whose single big number can't survive a second question.

Where TI fits

A range with a method behind it is worth more than a big round number. Takwimu Intelligence (TI) weights answers to KNBS census totals at the level where those totals exist, in practice the county, and shows each figure's margin of error, so a station can state its reach in a sentence a buyer can check. The Reach note for an advertiser sets it out.

To see it on your own station or title, book a walk-through.

Takwimu Intelligence (TI) is audience measurement software for Kenyan media.

Sources

  1. 1

    The RAJAR Methodology · RAJARThe hybrid survey since autumn 2021, MediaCell and the five-minute rule. Opened 5 November 2026.

  2. 2

    Margin of Sampling Error/Credibility Interval · American Association for Public Opinion ResearchWhen a margin of sampling error applies. Opened 5 November 2026.

  3. 3

    Democracy in media research: power of the markets (PAMRO 2022) · Ipsos, Barnabas Muya, 7 November 2022The Kenyan tracker design, barriers to measurement in sub-Saharan Africa, and joint industry committees. Opened 5 November 2026.

  4. 4

    Audience Measurement and Industry Trends Report, August 2026 (quarters one to four of 2025/26) · Communications Authority of KenyaSections 1, 2.1 to 2.7, 3.1, 3.2, 4.3, 4.4, 7 and Tables 1 to 8 (April to June 2026 is Q4 2025/26). The report prints a base of N=32.23M and states no sample size or method. Opened 5 November 2026.

More from the library

  • Streaming analytics: what a station's own server can count

    Radio streaming analytics count connections, sessions and hours on a station's stream. They are exact for the stream and silent about FM, so they can't give a station's reach.

  • Why on-air responses do not measure reach

    A radio call-in response rate needs the size of the audience as its denominator, and that is the number a station is trying to find. On-air replies count volunteers, and they can't give an audience size.

  • What is audience measurement?

    Audience measurement estimates how many people, and which people, were exposed to a station, channel, title or site in a stated period. It works from samples and panels, so every figure is an estimate.