Audience metrics are the numbers that show how many people a station, channel or title reached, how often it reached them, who they were and what each thousand cost. Every one of them is a fraction, so the first question to ask of any figure is what sits underneath it.
Key points:
- Reach counts each person once. Impressions count every exposure.
- A rating divides an audience by the whole population. Share divides it by the people using the medium at that moment.
- GRP, TRP and CPM are built from those parts, so an error in a rating carries straight into the price.
- Reach figures never add up across stations, surveys or platforms.
- A station needs three separate facts: who could receive it, who tuned in and what they thought.
What are audience metrics?
An audience metric turns behaviour into a number a buyer can compare. Sales teams use them to price airtime. Programme directors use them to judge a schedule. Advertisers use them to decide where the money goes.
The definitions in this section come from the trade's own glossaries: the Marketing Accountability Standards Board, Nielsen, RAJAR and Barb. The Kenyan figures come from the Communications Authority of Kenya (CA) and named research firms. Each page gives the formula, a worked example with invented round numbers, and the way the metric applies here.
Audience metrics at a glance
| Metric | The question it answers | Formula |
|---|---|---|
| Reach | How many different people were exposed at least once? | People reached ÷ universe × 100 |
| Reach and frequency | How often did those people get the message? | Impressions ÷ people reached |
| Share of audience | Of everyone using the medium, what percentage chose you? | Your audience ÷ total audience for the medium × 100 |
| Rating vs share | How does a rating differ from a share? | Rating = audience ÷ universe × 100 |
| Gross rating point | How heavy is the schedule? | Reach % × average frequency |
| TRP meaning | How heavy is it among the target? | Target impressions ÷ target audience × 100 |
| Cost per thousand | What does each thousand exposures cost? | Cost ÷ impressions × 1,000 |
| Weekly reach in radio | How does the week compare with an average day? | Daily reach × 7 ÷ weekly reach = days per listener |
| Audience profile | Who is in the audience? | Group's audience ÷ total audience × 100 |
| Audience retention in radio | How long do people stay? | Hours in period × average audience ÷ cume |
| Share of voice | How loud is a brand against rivals? | Brand impressions ÷ all brands' impressions × 100 |
One campaign, every metric
Here is one invented campaign with round numbers. A county has 1,000,000 adults. A station sells a KSh 600,000 schedule that delivers 1,200,000 impressions to 500,000 different people.
| Metric | Sum | Result |
|---|---|---|
| Reach | 500,000 ÷ 1,000,000 × 100 | 50% |
| Average frequency | 1,200,000 ÷ 500,000 | 2.4 |
| Gross rating points | 50 × 2.4 | 120 |
| Cost per rating point | KSh 600,000 ÷ 120 | KSh 5,000 |
| Cost per thousand | KSh 600,000 ÷ 1,200,000 × 1,000 | KSh 500 |
Now take one moment in that schedule. At 7am, 300,000 of those adults are listening to radio and 75,000 are on this station. Its rating is 75,000 ÷ 1,000,000 × 100 = 7.5, and its share is 75,000 ÷ 300,000 × 100 = 25%.
That is seven numbers from one campaign. Each answers a different question, and none can stand in for another.
Why the bottom number decides everything
The universe is the total population of the audience category being measured, as the Barb glossary puts it. Change the universe and every percentage moves.
In Kenya the natural starting point is the census: KNBS enumerated 47,564,296 people in 2019. A county station's universe is the adults in its counties, and a national chart uses the nation. A percentage with no stated universe cannot be compared with anything.
What Kenyan figures already show
The CA report of August 2026 put radio at 75% over seven days in Q4 2025/26, and daily radio use at 49%. Those are figures for the whole medium.
GeoPoll, for January to March 2026, gave Citizen Radio 64.7% reach over the quarter, and in weekday breakfast gave Classic 105 18.8% of Top 10 listening and Citizen 17.4%. RadioInfo Africa, on the Ipsos survey for October to December 2025, gave Citizen a 17% market share.
Different windows. Different bases. All of these can be correct at once, and the mistake is to place them on one chart.
Why reach figures never add up
RAJAR/Ipsos/RSMB gives the cleanest proof. For the period ending 14 December 2025, UK BBC radio reached 30,904,000 adults a week, commercial radio 39,431,000 and other radio 4,143,000. Those three sum to 74,478,000, yet All Radio reached 50,000,000.
Hours behave differently. The same three lines of total hours do sum to All Radio's total, and their shares of 42.8%, 55.2% and 2.0% sum to 100%. Time adds. People overlap.
How the metrics connect
Most of these numbers form one chain. Divide an audience by the universe and you have a rating. Add the ratings of every spot and you have GRPs. Divide cost by GRPs for cost per point, or divide cost by impressions for CPM.
The chain has a consequence. Every later figure inherits the first figure's error. If a rating is inflated by a third, so are the GRPs, and the CPM looks a quarter cheaper than it is.
Four sources that must stay apart
Stations hold several kinds of number, and each records something different.
- A recall survey asks a sample of people what they listened to, and can be weighted to a population.
- A streaming log counts sessions on the station's own server.
- An on-air response counts people who chose to text or call.
- An advert log records what ran and when.
Each is useful on its own line. Added together, they produce a number that describes nobody.
Which audience metric for which job?
| Who is asking | What they need to know | Start with |
|---|---|---|
| Programme director | Is this show winning its time slot? | Share, time spent, retention |
| Sales manager | How big is the audience, and who is in it? | Reach, rating, audience profile |
| Advertiser or planner | What did the schedule deliver for the money? | GRP, TRP, CPM |
| Owner or board | Is the station growing, and do people rate it? | Reach trend, appreciation |
What do practitioners get wrong?
The commonest fault is quoting a metric with no base. A share with no daypart, or a reach with no window, is a claim nobody can test. The second fault is treating the largest number available as the station's audience. Both are easy to fix.
Takwimu Intelligence (TI) takes one position across this library. Coverage, reach and appreciation are three separate facts. Report all three, state the base for each, and keep every source on its own line.
Read next
- Audience measurement software for Kenyan media: the TI pack page for this work
- What is reach in media?
- Rating vs share
- How media reach is measured in Kenya
- Takwimu Intelligence (TI) home page
- Audience measurement in Kenya: the guide to this library
- The TI platform and how we check our numbers
Where TI fits
Audience metrics are only as strong as the questions behind them. TI measures audiences by asking people: its day-after recall questionnaire produces station share, daily and weekly reach and reach by daypart, weighted to KNBS census totals at county level, with a margin of error on each figure. Its appreciation index sits beside them, because size and regard are different facts.
To see it on your own station or title, book a walk-through.