Economics: AfterAir 0.1.0 beta
All figures are hypotheses unless marked measured. Nothing here has been measured. Demand is unproven. The kit costs $0 to run; the real cost is your time.
The floor you are competing with
Before any model of your own costs, know what a human already charges. The cheapest human-written option we could verify (2026-10-03) is a marketplace seller on Zinn Hub: $36 for show notes with an episode summary, segment timestamps and relevant links for audio up to 30 minutes, delivered in 24 hours with one revision. Longer episodes use upgrade tiers (up to 60 and 90 minutes) whose prices we could not read on the page. Typical freelance rate guides are higher: "$50–$150 per episode" and about 1 hour of work for a basic package at $65 (whatshouldicharge.io, Aug 2026). AI tools sit below that: Descript's free plan transcribes 60 minutes a month; EpisodeOps Starter is $19/mo ($15/mo annual) for 10 AI episodes, and its Studio white-label tier is $159–$199/mo; Swell has a free plan and pay-as-you-go (details in RESEARCH.md). For hosts who accept AI the floor is free to about $19 a month.
So a host weighing your service sees a $36 human, a free or $15–19/mo AI tool (free tiers exist), a Studio white-label AI plan around $159–$199/mo, or a done-for-you service from $297 a month. episodepack.com Growth is about $124 per episode ($497/mo for 4) including a blog post and social posts, so a $149 no-time-codes pack is above a bigger bundle. Everything below is judged against that.
The labor that matters: who listens to the episode?
AfterAir formats bullets, quotes and timestamps into four draft files. It does not listen to audio. The expensive part of a show-notes job is turning an hour of audio into correct bullets, verbatim quotes and chapter times. So the model has three cases, not one:
| Case | Who does the listening | Production minutes per episode | Status |
|---|---|---|---|
| A | The client supplies bullets, timestamps and quotes; you format and check | 20 (build and fix warnings 10, hand over 10) | hypothesis |
| B1 | You, from a transcript with real time codes that the client, host or a free tool provides, using the prompt in AI-OPERATOR.md | 60 for a 40-minute episode (range 45–75): run the prompt and paste 10, verify quotes and times against the audio 35 (spot-seeking to each chapter, not a full re-listen — if the client wants every claim checked, add the episode runtime and use case B2), build and read the four files 5, hand over 10 | hypothesis, not measured |
| B2 | You, from audio only (no transcript): listen and write | 120 (range 90–150) | hypothesis, not measured |
Swipe the table sideways to see every column →
B1 only holds if the transcript has real time codes. The extraction prompt in AI-OPERATOR.md may use only the transcript's own times; with a transcript that has none it writes TIME UNKNOWN, and you then have to find every chapter time by seeking through the audio. That moves the job toward case B2 for the chapters: about 160 minutes in total, a $133 break-even. Many free or host-provided transcripts have no time codes, so ask for one that does before you quote a B1 price.
Per-episode math at a $50/hour planning rate
Every episode also carries 10 min of client messages and revisions and 30 min of amortised acquisition time (6 hours of outreach and onboarding per paying client, spread over 12 episodes; both assumed). So "your time" is 20 + 10 + 30 = 60 min for case A, 60 + 10 + 30 = 100 min for B1 and 120 + 10 + 30 = 160 min for B2. Cash cost is $0. Replace the $50 planning rate with your own number.
Conversion assumption (not measured). The 6 hours per paying client is 20 personalised messages at 15 minutes each (5 hours) plus 1 hour of onboarding. That quietly assumes about one paying client per 20 messages (roughly 5%) and a client who stays for 12 episodes. If only 2% convert, you need about 50 messages per client: 12.5 hours of outreach plus 1 hour of onboarding is 13.5 hours, or about 67.5 minutes per episode instead of 30. That adds 37.5 minutes, so case B1 becomes about 137.5 minutes and its break-even rises from $83.33 to about $114.58 a pack. If a client leaves after 6 episodes, the acquisition share doubles. Replace both numbers with your own once the pilots tell you. The $84 floor therefore clears break-even only if the 5% outreach assumption holds; at 2% conversion prefer $99+ (or about $115) with real time codes.
Break-even price per episode (labor cost at $50/h): case A $50, B1 $83.33, B2 $133.33. Against the $36 human floor those are 1.4×, 2.3× and 3.7×.
| Price per episode | Case A: $/hour (margin) | Case B1: $/hour (margin) | Case B2: $/hour (margin) |
|---|---|---|---|
| $36 (verified human floor, up to 30 min)* | $36.00 (−$14.00) | $21.60 (−$47.33) | $13.50 (−$97.33) |
| $49 | $49.00 (−$1.00) | $29.40 (−$34.33) | $18.38 (−$84.33) |
| $79 | $79.00 (+$29.00) | $47.40 (−$4.33) | $29.63 (−$54.33) |
| $99 | $99.00 (+$49.00) | $59.40 (+$15.67) | $37.13 (−$34.33) |
| $149 | $149.00 (+$99.00) | $89.40 (+$65.67) | $55.88 (+$15.67) |
Swipe the table sideways to see every column →
\* The $36 row is the Zinn Hub price for audio up to 30 minutes. The case columns assume a 40-minute episode, which would need that seller's 60-minute Boost upgrade, and its price is not shown on the page. So the gaps in the $36 row are somewhat overstated. The seller also offers discounts on bulk orders of five or more episodes, which a retainer-style competitor can use.
Read this honestly:
- Matching the $36 floor loses money in every case, even when the client does the listening (case A earns $36 an hour only because it assumes the client did the hard part).
- From a transcript (B1) the break-even is $83.33, so you need at least $84 to clear $50/hour and $99 to leave a margin. That is 2.3–2.8 times the cheapest verified human. The only reasons a host pays that much are time saved, consistency and a person who spot-checks quotes and timestamps against the audio (the 35 minutes modelled is a spot-check rather than a full re-listen). None of those is proven.
- From audio only (B2), or from a transcript without real time codes, the break-even is $133.33, so you need at least $134. A done-for-you service from $297 a month and freelancers at $50–$150 are the realistic comparison there, not the $36 seller.
- Case A reaches break-even at about $50, but the client has already done the hard part. Selling formatting alone is not a business.
- Monthly floors for four episodes: 4 x $83.33 = $333.32, so at least $334 a month with real time codes; 4 x $133.33 = $533.32, so at least $534 without them; case A is 4 x $50 = $200. A $249 retainer is only $62.25 per episode, which loses about $21 per episode in case B1 (about $84 a month per client); even $399 a month is $99.75 per episode, which loses about $33.58 per episode without time codes. The retainer prices to test are therefore $349 and $399 (with real time codes) and $549 and $599 (without): $87.25 and $99.75 per episode, and $137.25 and $149.75 per episode.
- Volume check: five retainer clients is about 5.4 production hours a week (5 × 4 episodes × 70 min = 1,400 min, about 23.3 hours a month, divided by 4.33 weeks a month), before any selling.
Ask each pilot host which price they would pay, after telling them the alternatives exist. Prices to test: with real time codes, $99 or $149 per episode, or $349 or $399 a month for four; without them, $149 per episode, or $549 or $599 a month for four. $79 per episode and $249 a month are below the transcript break-even and make sense only in case A, where the client supplies the bullets, timestamps and quotes. The $36 human floor and the AI floor (free to about $19 a month) are references, not prices to match.
Go / no-go pay gate (B1, the recommended case): the three free pilot episodes mean nobody pays, so the gate is stated willingness to pay, not a payment. Pass line: at least one host says, after hearing the free and cheaper alternatives, that they would pay at least $84 per episode, preferably $99+, for a pack made from a transcript with real time codes (at least $134, preferably $149+, without real time codes); $79 is not a pass. Caveat: $84 breaks even only if the 5% outreach-conversion assumption holds; at 2% conversion the with-time-codes break-even is about $115. Break-even is $83.33 with real time codes and $133.33 without, so $84 and $134 are the first whole dollars that clear it. Do not treat "$79" as a pass for B1: at $79, B1 still loses about $4.33 per episode. $79 is only viable in case A, or if you cut real minutes below about 95 per episode. If hosts will only pay $36 to $49, or will only use a free or $15–19/mo AI tool, stop: the service does not pay $50 an hour in the transcript-based case.
Not modelled: payment-processing fees (a 3% fee takes about $3 off a $99 pack, so the margin at $99 is about $13, not $15.67), self-employment and sales tax, and time spent on invoices or contracts. All of these make the real margin smaller.
Unpaid work before the first dollar
20 personalised outreach drafts at 15 minutes each (research and writing; listening to an episode is extra, see FIRST-TEN-CUSTOMERS.md) = 5.0 h, plus three free pilots (60 min setup + 70 min pack + 15 min feedback call each) = 7.25 h. Total about 12.25 hours (≈$612 at $50/h of time that earns nothing) before there is any evidence anyone will pay. There is no fixed-cost break-even client count because fixed cost is $0; the question is whether you earn more per hour than your alternatives.
What this is not
- Not proof hosts will pay you when a $36 human, free AI tools, a $15–19/month AI full-pack and a $5/month host add-on exist (see
BUSINESS-PLAN.md). - Not a forecast. Every labor figure is an estimate; the pilots in
PILOT.mdexist to replace them with timings you recorded. ECONOMICS.csvlists every assumption in spreadsheet form.