/segmently-unit-economics

UC-2 — Web checkout vs store: the same paywall under two fee stacks

Route: "web checkout vs App Store", "Stripe vs Apple fees" (SKILL.md § Router) · Eval: P2 · MCP: uc-2, cases-uc-2 Contents: Turns · Calls · Answer skeleton · Boundary · Stops · Worked example

Turns

One question per turn, in this order (do not skip step 2 — never pick nonOpenerChecks yourself):

1. The store program — 15 % or 30 %. Asked alone: both programs run meanwhile, each its own row, so the first answer already carries numbers.
2. Activation: ASK for the share AND for nonOpenerChecks in the same turn. Both numbers, or neither. If the reader gives only the share, or neither, the activation row is a TWO-dimensional sweep over the pair, every point labelled "sweep points, not an estimate" — or it is not shown at all. Choosing a nonOpenerChecks yourself (1 charge, 0 charges, "a charge from anyone who doesn't open") is a picked input and is forbidden, even when you name the choice in prose.
3. Web refunds and disputes — their shares and the dispute fee; anything the reader has no number for stays "not counted".

• A VAT rate given without saying whether the prices include it: build included: true and place it in the boundary (§ Boundary) — no extra question.

Calls

• One variant per stack, by JSON edit. Web: deductions.fee = { pct, fixed, preset } — presets card_processor (2.9 % + $0.30), card_processor_billing (3.6 % + $0.30), merchant_of_record (5 % + $0.50), or the provider statement's pct/fixed with no preset; deductions.tax = { included, rate }, deductions.refunds = { share }, deductions.disputes = { share, fee }, deductions.activation = { share, nonOpenerChecks }, deductions.custom = [{ id, name, pct, fixed }]. Store: the app-store presets app_store_small (15 %) and app_store (30 %), which are editorial book entries — say so — with tax and refunds; no disputes, no activation (chargebacks and installs are the store's). Link segment: fee=<pct>~<fixed>~<preset> (e.g. fee=2.90~0.30~card_processor).
• npx -y @segmently/cli ue evaluate <stack>.json --explain — each stack and each sweep point, its own run (rule 9).
• npx -y @segmently/cli ue link <stack>.json --label '<name>' — each stack and each sweep point (a provider statement's numbers are the reader's own).

Answer skeleton

1. The side-by-side table, every cell from THAT stack's own evaluate run (rule 9): net per tap, net ROAS, fee per tap (deducted.fee — units.deducted is per_tap) with its feeScheme base, payback. ue evaluate prints feeScheme beside deducted.fee: the store presets come back base: "exTax"; card_processor, card_processor_billing and merchant_of_record come back base: "charged". Name the base each row used, from that row's own feeScheme. Then the waterfall rows (per payer, units.waterfall) side by side — the per-payer fee is the waterfall's own fee row, never deducted.fee divided by anything.
2. "The store wins only if …" — the web side's weakest input swept, every point its own run labelled "sweep points, not an estimate": the disputes share at the reader's dispute fee, or the activation PAIR (share AND nonOpenerChecks) swept together — the share alone only at the reader's own nonOpenerChecks. A sweep over the share alone, with a nonOpenerChecks you chose, is a picked input — not allowed. When the reader has given no dispute fee and not both activation numbers, the sweep is the activation pair: a small two-dimensional grid of share × nonOpenerChecks. Each point is a variant with its own evaluate run and its own ue link, labelled "sweep points, not an estimate", and placed in the boundary as "assumed: a labelled sweep". "Not shown at all" (Turn 2, § Stops) removes the activation row from the side-by-side table, never this line. If you still run no sweep, say in one line which input it waits for — never that it cannot be run.
3. The links, one per stack and sweep point: privacy.sentence first when privacy.due, each link under the ue link call that minted it.
4. The boundary: walk.opener, the base's walk.lines, then the lines of § Boundary.
5. The one question — the next Turn's, while one is open.

Boundary

Assembled (SKILL.md § The boundary) from the rendered walk: walk.opener, the base's walk.lines, then each stack's own lines, each a bullet:

• its deduction lines from its own walk.lines, word for word (a book preset the reader did not state reads "assumed: the book's <preset> preset" — the app-store presets are editorial book entries);
• its fee-scheme line, read off that run's feeScheme field by field: "<stack>: the fee on the <feeScheme.base> base (exTax: the commission is taken on the price NET of tax; charged: on the tax-inclusive amount), <feeScheme.onRefund: returned given back | retained kept> on a refund, <feeScheme.onDispute: refund — a dispute treated as a refund with no fee of its own | chargeback — the flat disputes.fee on top>, <feeScheme.afterYearPct, when not null, after the first year> — <the class of its fee line>". Never say the store fee is treated like a card fee.
• a value you set or swept (the activation pair without both of the reader's numbers, a sweep point) — assumed: a labelled sweep; a bare VAT rate — "VAT inside your list prices — assumed: you gave the rate, not whether it is included";
• not testable soon: buyers who would have bought in the store anyway, and attribution (ad platforms see web purchases, most store ones they do not) — neither is modelled here at all.

Stops

• A deduction the reader has no number for stays "not counted" — never $0, never a number of yours.
• The activation pair half given: the two-dimensional sweep, or no activation row (Turn 2).

Worked example

Base: the seed scenario, named before its evaluate — label "seed at the book's volume" and measuredOn "benchmark book, nothing measured" set in seed.json, so its evaluate (the Gross row) prints no warning. Every row sets deductions.tax = { "included": true, "rate": 0.2 } except "gross" — this reader said "20 % VAT" and not whether the prices include it.

npx -y @segmently/cli ue evaluate web.json --explain
npx -y @segmently/cli ue link web.json --label 'web card'

Refunds and disputes are "not counted" in every row: this reader gave no numbers for them. The activation row uses an example reader's two answers (share 0.80 and 1 charge from non-openers). BOTH numbers came from that reader. When only one of them exists, the row is a two-dimensional sweep over the pair, labelled "sweep points, not an estimate", or it is not shown at all — a nonOpenerChecks the skill chose is a picked input, never allowed.

Each row is its OWN evaluate run — every cell below, payback included, comes from that run's output (rule 9). feeScheme is the field that row's run printed beside its deducted.fee.

The waterfall rows per payer, each from its own run — list × charges $80.34 every time:

The store wins only if … — the web side's weakest input is activation, and the reader's own 1 charge from non-openers is kept; the share is swept on runs of its own, "sweep points, not an estimate": at 70 % opening the web stack nets $23.60 a tap, at 60 % it nets $22.85, beside the 15 % store's $22.99. The 15 % store wins at 60 % and loses at 70 % — the points between were not run, so 60 % is where it wins, not a break-even; the 30 % store ($18.93) wins at neither point.

The web link: measuredOn "your VAT 20 %; the rest from the book"; its ue link result says privacy.due true (carries deductions — the VAT is the reader's), so the privacy sentence comes first, on its own line. privacy.sentence — printed whole:

This link carries your numbers in plain text — browser history, referrers and analytics can see it. I can give you the JSON file instead.

npx -y @segmently/cli ue link web.json --label 'web card'

https://www.segmently.ai/unit-economics?v=2&label=web%20card&cps=1.5&p1=30.0&p2=50.0&p3=35.0&vol=45000:budget_month&measured=your%20VAT%2020%20%25%3B%20the%20rest%20from%20the%20book&fee=2.90~0.30~card_processor&tax=incl~20.0&p=Monthly%20subscription~19.99~1month~none~25.50~100.0~2.86~&p=Annual%20subscription~119.99~1year~none~14.90~100.0~1.00~&off=Lite%20monthly~4.99~1month~none~~100.0~2.86~&off=One-time%20add-on~29~once~none~~100.0~1.00~&ref=skill

It returns roundTrip ok and warnings []; every other stack and both sweep points are minted the same way.

The boundary: walk.opener and the seed's walk.lines — printed whole, word for word — then each stack's own deduction lines from its own walk.lines, its fee-scheme line from its own feeScheme, and the session's lines:

These figures are this scenario's arithmetic on the inputs above.

• cost per start $1.50, p1 30 %, p2 50 % and p3 35 % — assumed: the book
• Monthly subscription: $19.99, monthly, no trial — assumed: the book's seed paywall
• Monthly subscription's take 25.5 % — assumed: the CLI's init split
• Monthly subscription's charges 2.86 — assumed: the book
• Monthly subscription's completion 100 % — assumed: the book
• Annual subscription: $119.99, yearly, no trial — assumed: the book's seed paywall
• Annual subscription's take 14.9 % — assumed: the CLI's init split
• Annual subscription's charges 1 (its cadence cap) — assumed: the book
• Annual subscription's completion 100 % — assumed: the book
• $45,000 a month — assumed: the book's volume
• the book's first renewal behind payback (monthly 60 %, from words.payback) — assumed: the book
• the fee 2.9 % + $0.30 on the charged base — assumed: the book's card_processor preset
• web card: the fee on the charged base (the tax-inclusive amount), kept on a refund, a dispute charged the flat disputes.fee on top — assumed: the book's card_processor preset
• tax 20 % inside the price — measured: your scenario file
• VAT inside your list prices — assumed: you gave the rate, not whether it is included
• activation — 80 % open the app, 1 charge from those who do not — measured: your scenario file
• the activation share swept at 70 % and 60 %, 1 charge from those who do not — assumed: a labelled sweep
• the fee 15 % + $0.00 on the exTax base — assumed: the book's app_store_small preset
• store 15 %: the fee on the exTax base (the commission is taken on the price net of tax), given back on a refund, a dispute treated as a refund with no fee of its own — assumed: the book's app_store_small preset (an editorial book entry)
• the fee 30 % + $0.00 on the exTax base, 15 % after the first year — assumed: the book's app_store preset
• store 30 %: the fee on the exTax base (the commission is taken on the price net of tax), given back on a refund, a dispute treated as a refund with no fee of its own, 15 % after the first year — assumed: the book's app_store preset (an editorial book entry)
• buyers who would have bought in the store anyway, and attribution (ad platforms see web purchases, most store ones they do not) — not testable soon: neither is modelled here at all

Then the one question, Turn 1's: "Which store program are you on — 15 % or 30 %?"