App Pricing Localize

app-pricing-localize

Compute an iOS subscription's per-country App Store prices. Computes PPP as the default baseline, optionally adds a competitor-benchmark and/or App Masters column, shown side by side (local + USD), snapped to real Apple price points. Produces three artifacts: a compare view, a single-product report, and — for multiple products — a wide matrix (one column per tier, current→new cells) that is the human-editable source of truth.

This skill is read-only against the stores. To push the resulting matrix/report live to the App Store or Google Play, hand it to the sibling app-pricing-apply skill. Works for any App Store app — configure via env vars (ASC_KEY_ID, ASC_ISSUER_ID, ASC_KEY_PATH, ASC_BUNDLE_ID) plus the product identifier(s) you want to price.

Question flow (ask these, in order)

1. Baseline currency and anchor price? Default USD, anchor = the product's current US price (pulled from ASC). Override with --anchor-currency / --anchor-price N. A non-USD baseline is converted to USD internally for the PPP math.
2. Competitor benchmark? Yes / no.

• Yes: name a competitor (or pass an existing benchmark file). The skill will orchestrate research → tier listing → user pick (see the four-step flow below).
• Skip: PPP-only (plus optional App Masters).

1. Use the App Masters approach? Optional yes/no. App Masters = its revenue-tested charm ladder + same-number anchor for near-parity currencies (A$/€/C$ = the anchor number).

Four-step flow (when benchmark is on)

1. Generate the benchmark file — invoke app-pricing-research for the competitor + the user's country scope. Save to /tmp/.md (or reuse an existing file). Its per-tier tables (Local price + ≈USD, one section per tier) ARE the tier menu; there is no separate "list-tiers" command — read the raw file directly.

1. Ask the user how to use those tiers. Two options:

• (1) Pick a specific tier by name — e.g. "ChatGPT Plus (1)" (monthly), "ChatGPT Go", etc. The BENCHMARK column will scale by that tier's per-country ratios (price_country / price_US), with USA anchored at your anchor. Use when the user has a clear positioning call ("we're competing with X monthly").
• (2) Use lowest — for each country, take the lowest ratio across ALL tiers (the biggest regional discount the competitor gives in that country, from whichever tier). Most aggressive benchmark; useful for "match the deepest discount competitors are willing to give per market" strategy.

1. compare --benchmark-file --benchmark-tier '' --approaches ppp,benchmark,appmasters — side-by-side comparison. Under option (2) the BENCHMARK column can mix tiers per country (India might follow Go's ratio, Germany might follow Plus's).

1. Ask which approach (ppp / benchmark / appmasters), then report --approach X — writes one markdown file with ## Comparison, ## Final pricing — X, and ## Apply (asc commands). Shareable artifact; nothing is written to App Store Connect.

Skip to step 3 (no benchmark)

If the user skips benchmark, go straight to compare --approaches ppp[,appmasters] → pick approach → report. --benchmark-tier is only required when --benchmark-file is passed AND benchmark is in --approaches.

Quick start

# One-time: point at your ASC key (any App Store app you have Admin/App-Manager access to).
export ASC_KEY_ID=XXXXXXXXXX
export ASC_ISSUER_ID=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
export ASC_KEY_PATH=~/.appstoreconnect/private_keys/AuthKey_XXXXXXXXXX.p8
export ASC_BUNDLE_ID=com.example.yourapp
PRODUCT=com.example.yourapp.monthly     # the subscription's productId
COUNTRIES=USA,IND,IDN,BRA,JPN,KOR,MEX,PHL,DEU,GBR,CAN,AUS,TWN

cd skills/app-pricing-localize/scripts

# 1. (Skip if you have a file) Generate a competitor benchmark via the sibling skill.
python3 ../../app-pricing-research/scripts/benchmark.py --app-id 6448311069 \
  --countries us,in,id,br,jp,kr,mx,ph,de,gb,ca,au,tw --out /tmp/chatgpt.md

# 2. Read /tmp/chatgpt.md to see the tier names.
#    Ask the user: pick a tier by name OR use 'lowest' (per-country lowest ratio).

# 3. Comparison (all three approaches). Named tier:
python3 localize.py compare --product $PRODUCT \
  --approaches ppp,benchmark,appmasters \
  --benchmark-file /tmp/chatgpt.md --benchmark-tier 'ChatGPT Plus (1)' \
  --countries $COUNTRIES

# 3b. Same but with 'lowest' — per-country deepest discount across every tier in the file.
python3 localize.py compare --product $PRODUCT \
  --approaches ppp,benchmark,appmasters \
  --benchmark-file /tmp/chatgpt.md --benchmark-tier lowest \
  --countries $COUNTRIES

# 4. After the user picks an approach — bundle comparison + final table + asc commands into ONE md.
python3 localize.py report --product $PRODUCT \
  --approach ppp --approaches ppp,benchmark,appmasters \
  --benchmark-file /tmp/chatgpt.md --benchmark-tier 'ChatGPT Plus (1)' \
  --countries $COUNTRIES --out /tmp/pricing-plan.md

Knobs

• --anchor-currency USD (default) · --anchor-price N (default = current US price of --product).
• --k 0.6 — share of the PPP gap to take (0 = no discount, 1 = full PPP). --floor 0.40 — never below 40% of the anchor. Both are tunable defaults, not constants — calibrate against your own elasticity data.
• --approaches ppp[,benchmark,appmasters] — columns to render in the comparison · --approach ppp|benchmark|appmasters — the user's chosen one (required for report; auto-folded into --approaches if missing) · --countries iso3 list · --out file.
• matrix knobs: --products P1,P2,… (one column each) · --labels 'Weekly,Monthly,…' (friendly column headers, parallel to --products; these are the tier names app-pricing-apply matches with --tier) · --usd-ref N (which product feeds the ≈$ PPP-depth column; default 0) · plus --k --floor --countries --out.

Matrix (multiple products → one table)

For a whole subscription lineup, matrix computes PPP for several products and merges them into one wide table — one column per tier, current→new cells, USA/anchor row first, poorest→richest. This .md is the editable source of truth: tweak any price by hand, then apply it.

python3 localize.py matrix \
  --products com.example.weekly,com.example.monthly,com.example.yearly \
  --labels 'Weekly,Monthly,Yearly' --usd-ref 1 --out matrix.md

Applying (App Store / Google Play)

This skill never writes to a store. Hand the matrix .md (or a report) to the sibling app-pricing-apply skill, which applies one tier at a time to the App Store (scheduled price changes via the ASC API) and/or Google Play (regional configs via the Android Publisher API) — dry-run by default, only ever lowering. After you produce a matrix, offer it: "matrix ready — want me to apply it to the App Store / Google Play? (app-pricing-apply)". That skill documents the store realities (Apple can't preserve existing subscribers on a decrease; Google reprices new-only; the ~2-day startDate lead; per-region currency handling).

How each column is computed

• PPP: target_usd = anchor × clamp(1 − k·(1 − price_level), floor, 1), then × FX, charm-snap, capped at current. Price levels = World Bank PA.NUS.PPP ÷ PA.NUS.FCRF (Taiwan/HK use a flagged estimate). FX = open.er-api.com.
• benchmark: the picked tier's per-country ratios, applied to your anchor. lowest = per-country min across all tiers.
• appmasters: the App Masters ladder interpolated at the anchor tier; for near-parity currencies not in the sheet, the same-number anchor. Sheet data in data/appmasters_ladder.csv (only 7 countries captured — extend if you get the full sheet).

Charm-snapping rule

The proposed price preserves the current price's last-two-digits tail where possible:

• Current ends in 99 (₹1299) or .99 ($12.99) → proposal ends in 99 / .99.
• Current ends in .90 (R$79.90) or 90 (NT$390) → proposal ends in .90 / 90.
• Current is round (¥2000, ₩19000, Rp229000) → magnitude-snap (¥1700, ₩15000, Rp130000).
• No current price known → magnitude-snap (safe default).

Notes / limits

• Coverage: every territory with World Bank data (≈ all). The report header prints "N of M priced territories".
• Staleness: World Bank PPP lags ~1–2 yr; FX is live. The header stamps both — re-check if old.
• Tax: EU/UK list VAT-inclusive prices, so a PPP target there is pre-tax-equivalent.
• The strategy (which approach, k, which benchmark) is a human call; this skill automates the mechanics + safe orchestration.

Prereqs

• ASC API key (same .p8 used for IPA uploads); chmod 600. python3 + PyJWT (pip install pyjwt cryptography).
• Benchmark column is optional. If you already have a benchmark markdown file (any source — a saved app-pricing-research report, an export, or hand-rolled), just pass --benchmark-file. Only invoke the app-pricing-research skill when you actually need to generate one.
• Setting prices needs an ASC-writing CLI (e.g. asc) — this skill only prints the commands.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: vicoa-ai
• Source: vicoa-ai/app-business-skills
• License: MIT

Install and usage instructions live in the source repository linked above.