Why audio ads need their own numbers
If you buy ads on music-streaming apps and internet radio, you are buying something unusual: a message that people hear but cannot see. The listener has a phone in a pocket, a speaker in the kitchen or a car system on the motorway. The ad plays between two songs or two programme blocks, usually cannot be skipped and can almost never be clicked. Most of the metrics you know from display advertising, clicks and click-through rates above all, tell you very little here.
That does not make audio unmeasurable. It means that the measurements are different. The app can report how many times the spot started and how many times it played to the very last second. Your own shop can count how often a spoken promo code is used. A short survey can ask people how they heard about you. And on the rare occasion a screen is on, a companion banner next to the audio can be clicked. Put together, these numbers describe a campaign well, but only if you divide the right figure by the right base.
This lesson covers the first and most important step: the listen-through rate, the eCPM and the cost per completed listen. The example is Dunlin Trail Co., a fictional online shop for trail-running shoes. All companies, apps, campaigns and numbers in this article and in the course are invented for learning, and every benchmark is illustrative. This article is about audio ads on music-streaming and radio apps, not about podcasts, which have their own measurement.
The core concepts: ad start, completed listen and listen-through rate
Three words carry the whole lesson. An ad start (many reports call it an impression) is counted when the spot begins to play. A completed listen is counted when the spot plays to its end. The listen-through rate, LTR for short, connects them: LTR = completed listens ÷ ad starts.
Why do some listeners not reach the end? The app is closed, the station is switched, the connection drops, a smart speaker is muted, or the phone goes to sleep. Because most audio spots cannot be skipped, the LTR is usually high, often in the nineties, and that makes small differences meaningful. A line that sits at 81 % while the others sit at 93 % is telling you something about the placement, the app or the creative.
The end of the spot is where the offer and the promo code are usually spoken, so a completed listen is the closest thing to a delivered message that an audio report offers. That is why the course treats the completed listen, not the impression, as the unit of value. The price of an ad start is still how audio is bought, though, and that brings us to the second pair of metrics.
Two prices: eCPM and the cost per completed listen
Audio is mostly traded per 1,000 ad starts. The eCPM (effective cost per mille) is spend ÷ ad starts × 1,000. It is the price of one thousand plays, and it is how sellers quote and buyers compare. What you actually want to know, however, is what a heard message costs: cost per completed listen = spend ÷ completed listens.
This number is a fraction of a euro, which is awkward to read, so the course shows it in euro cents (ct) by multiplying by 100. A price of 1.50 ct per completed listen means that one thousand completed listens cost €15. The two prices are linked by the LTR: cost per completed listen = eCPM ÷ LTR ÷ 1,000 in euros, or eCPM ÷ LTR ÷ 10 in cents. A weak LTR therefore raises the price of every message that is actually heard.
A fully worked example: Dunlin Trail Co. in a two-week spring campaign
Dunlin Trail Co. bought two spot lengths, a 30-second spot and a 15-second cut of the same Trailhead spot, plus a 30-second founder-voice version, on four fictional apps: Tunewave and Beatloft (music streaming), RadioNest (internet radio) and AirDial (a live-radio simulcast app). The delivery report has ten ad lines. These are the same ten lines you work with in the lab above.
| Ad line (app · creative) | Ad starts | Completed listens | Spend € | Listen-through rate | eCPM € | Cost per completed listen (ct) |
|---|---|---|---|---|---|---|
| Tunewave · 30s Trailhead | 184,000 | 171,900 | €2,576 | 93.4 % | 14.00 | 1.50 |
| Tunewave · 15s Trailhead cut | 126,000 | 121,200 | €1,512 | 96.2 % | 12.00 | 1.25 |
| RadioNest · 30s Trailhead | 96,500 | 84,300 | €1,158 | 87.4 % | 12.00 | 1.37 |
| RadioNest · 15s Trailhead cut | 72,000 | 66,800 | €792 | 92.8 % | 11.00 | 1.19 |
| Beatloft · 30s Trailhead | 143,000 | 136,400 | €2,288 | 95.4 % | 16.00 | 1.68 |
| Beatloft · 15s Trailhead cut | 88,000 | 85,100 | €1,232 | 96.7 % | 14.00 | 1.45 |
| AirDial · 30s Trailhead | 64,000 | 51,800 | €608 | 80.9 % | 9.50 | 1.17 |
| AirDial · 15s Trailhead cut | 41,500 | 36,100 | €353 | 87.0 % | 8.51 | 0.98 |
| Tunewave · 30s Founder voice | 110,000 | 104,700 | €1,650 | 95.2 % | 15.00 | 1.58 |
| RadioNest · 30s Founder voice | 58,000 | 52,300 | €725 | 90.2 % | 12.50 | 1.39 |
| Total | 983,000 | 910,600 | €12,894 | 92.6 % | 13.12 | 1.42 |
The formulas, step by step
Everything in the table comes from three columns of the delivery report (ad starts, completed listens, spend) and three short formulas. Here they are for row 4 of the lab, copied down to row 13:
- Listen-through rate (E4):
=C4/B4, formatted as a percentage. Completed listens divided by ad starts of the same line. - eCPM (F4):
=D4/B4*1000. Spend divided by ad starts, multiplied by 1,000 for the price of a thousand plays. - Cost per completed listen in cents (G4):
=D4/C4*100. Spend divided by completed listens, multiplied by 100 to show cents instead of fractions of a euro. - Campaign rate (E14):
=C14/B14, using the totals in row 14, not the average of the ten rates above it.
Take the first line, Tunewave 30s Trailhead. It had 184,000 ad starts and 171,900 completed listens, so the LTR is 171,900 ÷ 184,000 = 93.4 %. The spend was €2,576, so the eCPM is 2,576 ÷ 184,000 × 1,000 = €14.00, and the cost per completed listen is 2,576 ÷ 171,900 × 100 = 1.50 ct.
How to read the result
Start with the campaign level. Over the two weeks the ten lines delivered 983,000 ad starts, of which 910,600 were completed: a campaign LTR of 92.6 %. The campaign eCPM was €13.12 and the cost per completed listen 1.42 ct, so 72,400 ad starts, about 7.4 % of what was bought, were lost before the end of the spot.
Now look at the spread. The best line is Beatloft 15s Trailhead cut at 96.7 %, and the 15-second cuts are better than their 30-second parents in each app: shorter spots simply have fewer seconds in which a listener can leave. The weakest line is AirDial 30s Trailhead at 80.9 %. It also has the lowest eCPM of the 30-second lines, €9.50, which looks like a bargain until you notice that almost one in five plays does not reach the end.
The link between the two prices shows what a weak LTR costs. With an eCPM of €9.50 and an LTR of 80.9 %, AirDial 30s costs 9.50 ÷ 0.809 ÷ 10 = 1.17 ct per completed listen. If the same placement had reached the 93.4 % of Tunewave 30s, it would have cost 9.50 ÷ 0.934 ÷ 10 = 1.02 ct. The weak completion adds about 0.15 ct, roughly 15 %, to the price of every message that is actually heard. A cheap eCPM does not guarantee a cheap listen.
Why you never average the rates
The last row of the lab is a trap on purpose. If you click on the ten rates in column E and ask for their average, you get 91.5 %. The correct campaign rate, total completed listens divided by total ad starts, is 92.6 %. The gap of 1.1 points looks small, but it is systematic: the simple average treats AirDial 15s, which had 41,500 ad starts, exactly like Tunewave 30s with 184,000.
The rule is general and the course repeats it in every lesson: to combine rates, add the numerators and the denominators first, then divide. The same applies to the eCPM and to the cost per completed listen. A weighted ratio is what the whole campaign experienced; an unweighted average is what a statistician calls a different question.
Common mistakes and how to spot them
- Upside-down division.
=B4/C4gives a number above 1 (for example 1.07), which is impossible for a rate. If a listen-through rate shows more than 100 %, swap numerator and denominator. - Dividing by the wrong base.
=C4/$B$14divides one line by the starts of the whole campaign and returns tiny numbers like 17 %. Each line needs its own ad starts. - Mixing up the two prices. The eCPM divides by ad starts, the cost per completed listen divides by completed listens. If both columns show the same number, one of them uses the wrong column.
- Forgetting the factor.
=D4/B4is the price of one ad start (about 0.014 €). Multiply by 1,000 for the eCPM, or by 100 for cents per listen. - Averaging rates.
=AVERAGE(E4:E13)weights every line equally. Use=C14/B14instead. - Comparing across spot lengths without thought. A 15-second spot almost always has a higher LTR than a 30-second spot. Compare like with like, or compare the seconds heard, which lesson 2 teaches.
Quick reference: formulas in this lesson
| Measure | Formula in words | Example |
|---|---|---|
| Listen-through rate | Completed listens ÷ ad starts | =C4/B4 |
| eCPM | Spend ÷ ad starts × 1,000 | =D4/B4*1000 |
| Cost per completed listen (ct) | Spend ÷ completed listens × 100 | =D4/C4*100 |
| Campaign rate | Total completed ÷ total starts | =C14/B14 |
| Totals | Add a column | =SUM(B4:B13) |
Real-world uses
- Weekly delivery review. Paste the report of each app into the sheet and sort by the listen-through rate. The lines at the bottom are the first to question with the seller.
- Buying decisions. Compare offers on the cost per completed listen instead of the eCPM, so that a cheap placement with poor completion does not win by mistake.
- Creative choice. The same spot in 15 and 30 seconds, or with different voices, can be compared line by line. The completed listen is the first test a creative must pass.
- Reporting to management. A campaign LTR, an eCPM and a cost per completed listen fit on one line and are honest about what was bought.
- Setting a floor. Many teams agree a minimum listen-through rate with the seller and ask for a make-good when a line stays below it.
What you will build in the rest of this course
Lesson 1 is free. Lessons 2 to 20 are part of the full course, and each one has its own live lab with a new data set from the same fictional company. They build step by step from the delivery report to promo codes, surveys, tests and a final dashboard:
- Lesson 2, Quartile Reports: where listeners leave the spot: retention at 25, 50 and 75 percent, early drop-off and the seconds heard per ad start, with a retention chart.
- Lesson 3, Companion Banners: render rate, companion CTR and clicks per 1,000 audio impressions, split by device with SUMIFS.
- Lesson 4, Reach and Frequency: unique listeners, average frequency, reach per listening segment and a frequency-cap check with IF and COUNTIF.
- Lesson 5, Frequency Response: response rate by number of spots heard, the plateau, buyers per 1,000 impressions and a column chart.
- Lesson 6, Daypart Mix: share of ad starts against share of orders, an order index and the cost per order by time of day.
- Lesson 7, Promo-Code Response: redemptions per 1,000 listens, discount cost looked up with XLOOKUP, net revenue and net return per euro.
- Lesson 8, The Cost Ladder: cost per listen, per click and per redemption by app, with RANK.EQ and a bar chart.
- Lesson 9, Survey Response: response rate, the audio share of "how did you hear about us" answers and on-air against off-air weeks.
- Lesson 10, Brand-Lift Test: awareness lift in points, relative lift, z-score, p-value and a significance verdict.
- Lesson 11, Daypart × Weekday Grid: cost per order for 35 slots, weekday against weekend, INDEX/MATCH and a heat map.
- Lesson 12, Frequency-Cap What-If: the cost of every extra play, a keep-or-cut rule and the savings of a lower cap.
- Lesson 13, Creative Wear-Out: an index of weekly response, SLOPE, FORECAST.LINEAR and a replacement rule.
- Lesson 14, Redemption Lag: days from spot to order, an attribution window, WEEKDAY and the median lag.
- Lesson 15, Pacing Check: flight days, expected delivery, pacing and a Behind, On track or Ahead status from dates.
- Lesson 16, Data Quality: impossible and implausible report lines, flag counts and conditional formatting for duplicates.
- Lesson 17, Budget Reallocation: a diminishing-returns scenario with elasticities, projected redemptions and a comparison chart.
- Lesson 18, Flight Forecast: slope, intercept and R squared of spend against redemptions and a forecast for three budgets.
- Lesson 19, A/B Test Planning: z-values, units needed per arm, days needed and a check against the test window.
- Lesson 20, Final Project: a campaign dashboard with app totals, a chart and a highlight rule.
Summary
Audio ads are bought per 1,000 ad starts but valued by the messages that are heard to the end. The listen-through rate is =C4/B4, the eCPM is =D4/B4*1000 and the cost per completed listen is =D4/C4*100 cents. At Dunlin Trail Co. the ten ad lines delivered a campaign listen-through rate of 92.6 % at 1.42 ct per completed listen, with three lines under 88 % whose weak completion makes them cost more than their eCPM suggests. Always combine rates by dividing totals by totals.
Now scroll up to the lab and build the sheet yourself. It takes about fifteen minutes, and the Coach checks every cell as you go.
Frequently asked questions
What is the listen-through rate of an audio ad?
The listen-through rate is the share of ad starts that played to the end: completed listens divided by ad starts. Because most audio spots cannot be skipped, it is usually high, and a line that is clearly lower than the others points to a weak placement, app or creative.
How do I calculate the cost per completed listen in Excel?
Divide the spend by the completed listens and, because the result is a fraction of a euro, multiply by 100 to show cents: =D4/C4*100. It is more useful than the eCPM when lines differ in listen-through rate.
What is the difference between impressions, ad starts and completed listens?
An impression or ad start is counted when the spot begins to play. A completed listen is counted only when it plays to the last second. The gap between the two is lost to closed apps, switched stations, dropped connections and muted speakers.
Why should I not average the listen-through rates of my ad lines?
A simple average gives a small line the same weight as a large one. Add up the completed listens and the ad starts of all lines first and divide the totals, so that each line counts with its real size.
Is a 15-second audio spot better than a 30-second spot?
A 15-second spot usually has a higher listen-through rate, because there is less time to leave. That does not make it better: a longer spot can carry more of the message. Compare like with like or measure the seconds heard per ad start.
Does this course cover podcast advertising?
No. The course is about audio ads on music-streaming and radio apps. It covers listen-through rate, companion banners, frequency, dayparts, cost per listen, promo codes and surveys. Podcast ads have their own measurement and are not included.
Keep going: Streaming Audio Ads: Listen-Through Rate, Companion Banners, Frequency, Dayparts, Cost per Listen & Promo-Code Response
This free lab is lesson 1 of 20. Unlock the other 19 hands-on labs, the final exam and your certificate for a one-time $9.90.
- Quartile Reports: Retention at 25, 50 and 75 Percent, Early Drop-Off, Seconds Heard per Start and a Retention Chart
- Companion Banners: Render Rate, Companion CTR, Clicks per 1,000 Audio Impressions and the Device Split
- Reach and Frequency: Unique Listeners, Average Frequency, Segment Reach and a Frequency-Cap Check
- Frequency Response: Response Rate by Spots Heard, the Plateau, Buyers per 1,000 Impressions and Heavy-Listener Waste
- Daypart Mix: Share of Ad Starts, Share of Orders, the Order Index and Cost per Order by Time of Day
- Promo-Code Response: Redemptions per 1,000 Listens, Discount Cost, Net Revenue and Net Return per Euro
- The Cost Ladder: Cost per Listen, per Click and per Redemption by App, with Ranks and a Bar Chart
- Survey Response: Response Rate, the Audio Share of "How Did You Hear About Us" Answers and On-Air versus Off-Air Weeks
- Brand-Lift Test: Awareness Lift in Points, Relative Lift, z-Score, p-Value and a Significance Verdict
- The Daypart × Weekday Grid: Cost per Order for 35 Slots, Weekday versus Weekend, INDEX/MATCH and a Heat Map
- Frequency-Cap What-If: Cost of Each Extra Play, Cost per Extra Buyer, a Keep or Cut Rule and the Savings
- Creative Wear-Out: Weekly Response per 1,000 Ad Starts, an Index, the Weekly Slope, a Forecast and a Replacement Rule
- Redemption Lag: Days from Spot to Order, an Attribution Window, Weekday Numbers and the Median Lag
- Pacing Check: Flight Days, Days Elapsed, Expected Delivery, Pacing and Behind, On Track or Ahead Status
- Data Quality: Impossible Completions, Implausible Redemption Rates, Flag Counts and Duplicate Report Lines
- Budget Reallocation: Diminishing Returns with an Elasticity, Projected Redemptions, Cost per Redemption Now versus Proposed
- Flight Forecast: Slope, Intercept and R Squared of Spend versus Redemptions, FORECAST.LINEAR and Forecast Cost per Redemption
- A/B Test Planning: z-Values, Relative Uplift, Required Sample per Arm, Days Needed and a Fit-the-Window Check
- Final Project: The Spring Audio Campaign Dashboard with Listen-Through Rate, Cost per Redemption by App, a Chart and a Highlight Rule