---
name: "skill-retail-strategy"
type: skill
description: >-
  Set or refresh a retail venture's business strategy — category and
  positioning, range architecture, price architecture, channel sequence by
  stage, and the retail calendar — as the retail layer beneath
  skill-strategy-cascade. Use this when a retail venture (BARA, Keroma, any
  future store) needs its strategy set before the cascade can run, or when an
  operator asks for range, price or channel direction. Reads convention-retail
  for the channel model and KPI vocabulary and the venture's convention-brand
  for positioning, voice and discount posture. Output is a strategy statement
  plus the inputs skill-strategy-cascade needs, and the calendar spine that
  skill-gtm-plan and skill-launch-campaign plan against. Propose-only: Atlas
  drafts for an operator, states known against assumed, never invents the
  goal, never bulk-creates; a pivot is an Owner anchor; Aled approves.
  Surfaced by APP-1080; first draft 2026-09-03.
license: CC BY-NC-SA 4.0
---

# retail-strategy

Atlas's retail layer. `skill-strategy-charter` establishes why a venture exists and whom it serves; `skill-strategy-cascade` turns a stated goal into objectives and a KPI tree. Between them, for a retail venture, sits a set of decisions neither makes: what the range is, what it costs, where it sells, and when the year peaks. Without them the cascade sets objectives against a business nobody has described, which is how BARA and Keroma came to decide range, price and channel ad hoc. This skill proposes those decisions; it replaces neither neighbour. It reads `convention-retail` for the channel model and KPI vocabulary, `convention-brand` for positioning and voice, and `convention-shopify` for the KPI queries and the human-action boundary. Linear connector, Shopify read-only. Surfaced by APP-1080; first draft 2026-09-03.

## Trigger

- A retail venture needs its strategy set or refreshed before `skill-strategy-cascade` can run: the goal is stated, but range, price and channel are being decided ad hoc.
- An operator (`operator-bara`, or Keroma's when it gains one) asks for range, price or channel direction.
- A `task-strategy-review` gap ticket names a retail venture whose cascade cannot be baselined because this layer is missing.

Retail ventures only: BARA, Keroma, and any store that follows. For careerOS, A1 or Mr Pritchard the cascade runs without this layer.

## Behaviour

Five passes of analysis, then a proposal. Nothing is written to Linear until Pass 6, and nothing is created in bulk at all.

### Pass 1 — Read the picture, mark known against assumed

Read the charter and the stated goal (never inferred; if none exists, hand to `skill-strategy-charter` first), the venture's `convention-brand` instance and voice reference, its plan material (`bara.relaunch-plan`, `bara.road-to-seed`, the Keroma brief), and, where a store is live, the KPI baseline via the `convention-shopify` queries. Then draw one line through everything: what is known (a signed voice reference, a live line list, ninety days of orders) and what is assumed (an untested price band, a peer set chosen by taste). Every later pass carries that marking forward. A venture with no orders yet has almost everything assumed; say so.

### Pass 2 — Category and positioning

Draw the market map: the three to five brands the venture sits beside in a customer's head, and where it differs from each. For BARA the map is TOAST, Aesop and Ffern, and the difference is Welsh craft at the core with deeper maker involvement, as `convention-brand` records. For Keroma the map is streetwear-adjacent apparel; the peer set is assumed until the operator names it. From the map set the price band (what a customer expects to pay in this company) and the one reason to buy, a sentence a customer could repeat to a friend. Positioning must agree with the venture's `convention-brand` entry; where this pass would move it, that is a pivot (see Guardrails), not a strategy output.

### Pass 3 — Range architecture

Propose the range in five tiers, each with a job:

- **Hero** — the one or two lines the brand is known for; carries the story and the imagery.
- **Core** — the repeatable lines that make most of the revenue; restocked, not reinvented.
- **Entry** — the low-commitment first purchase that brings a new customer inside the price band.
- **Gift** — sets and giftable formats, priced and packaged for someone buying for someone else.
- **Limited** — dated, small-batch or collaborative lines that create real scarcity.

A 0→1 store carries few lines: as a working rule, eight to fifteen, every tier represented, hero and core filled first. More than that before there is sell-through data is inventory risk dressed as choice. Collection logic follows the tiers and the customer's ritual (the table, scent, washing up for BARA), not the maker's process. Thematic collections, whether a seasonal edit, a collaboration or a drop, sit under the parent brand with their own name and window but the parent's voice, palette and price band, so they add a chapter rather than start a second brand. A theme that needs its own voice is diluting, not extending.

### Pass 4 — Price architecture

Anchor pricing first: the hero line sets the ceiling the customer reads everything else against, so it is priced for what it is, not what it costs. Then the price ladder: a few steps between entry and hero, each a legible reason to trade up. Then bundles and sets: a modest saving on the parts or a gift format, never the only route to an acceptable AOV. Margin and the unit test belong to `skill-unit-economics`; this pass sets the architecture and hands the numbers there to check.

Discount posture is a brand decision and follows the venture's voice reference. It is one of two: real dated scarcity (a scheduled drop, a limited run, an end date that is true) or never-discount. Manufactured urgency, evergreen sales and countdown timers are off-voice for BARA (`bara.voice-reference`, approved 2026-08-10) and the default exclusion for any premium-craft venture here. Keroma's posture is unset until its brand instance exists; propose one and mark it assumed.

### Pass 5 — Channel sequence by stage, and the retail calendar

Sequence channels by the venture's stage, drawing the case for each from the channel model in `convention-retail` rather than restating it. At 0→1 that means the DTC store plus one or two owned channels the audience already uses; a further channel only once the first has sell-through data; marketplaces, wholesale and physical earning their place at scale, not at launch. Name each channel's job and the KPI it moves, in `convention-retail` vocabulary, so `skill-gtm-plan` can plan against it and `skill-channel-ops` can run it.

Then set the retail calendar, the spine `skill-gtm-plan` and `skill-launch-campaign` plan against: the gifting peaks that matter for this venture (Q4, Mother's Day, Father's Day, and for BARA Dydd Santes Dwynwen), the drop cadence for limited lines, the restock rhythm for core, and the quiet windows kept for building rather than selling. A peak the venture has never traded through is assumed, and marked so.

### Pass 6 — Propose, then hand to the cascade

Post the strategy for the operator and Aled so it reads in one pass: the strategy statement (one sentence on how this venture wins in its category), then positioning, range, price, channel sequence and calendar, each with its known or assumed marking. Alongside, list the inputs `skill-strategy-cascade` needs: the objectives the retail layer implies (hero sell-through, repeat rate, AOV, channel mix), each with a baseline or the source that will establish one, and the candidate KPIs under their `convention-retail` names. @mention Aled and assign him. On approval, file the strategy to the venture's project as a document and hand the inputs to the cascade, which drafts the KPI tree and takes Aled's targets under its own gate.

## Feeds

`skill-strategy-cascade` (objectives, baselines, candidate KPIs); `skill-gtm-plan` and `skill-launch-campaign` (calendar and channel sequence); `skill-unit-economics` (price architecture to test); `skill-storefront-merchandising` (range tiers and collection logic); `skill-audience-icp` (positioning and price band).

## Guardrails

- Never invents the goal. Works from a stated goal; if there is none, `skill-strategy-charter` runs first.
- Never bulk-creates. Proposes; Aled approves; the cascade creates initiatives under its own gate.
- States known against assumed on every output. An assumption dressed as fact is the failure mode.
- A pivot is an Owner anchor. Any change to positioning, category or price band is raised to the Owner with a recommendation, never actioned as a strategy output (`operator-bara`, Boundaries).
- Discount posture never contradicts the venture's voice reference.
- Prices and discounts are the operator's hand to release (`convention-shopify`); this skill designs them and stages nothing.
- Linear writes and read-only Shopify; never writes to the store, code or the delivery loop; never touches the Pipeline team.

## Setup

- Run on demand, or reopened from a `task-strategy-review` gap ticket. Connector: Linear, plus Shopify read-only for baselines.
- Not scheduled. Strategy is set deliberately, not on a clock.

Owned by Atlas (ATL-4S), run for an operator.

On finish, run skill-ops-retro capture on this run's friction: **search the Pulse queue first and add the evidence to a same-root note if one exists, otherwise file a new FRICTION note** for the drain pass to assess, in capture step 3's literal shape — team Apps, project os.Claude, `FRICTION:` title, labels `task` + `work:configuration` + `PUL-5E`, Backlog, unassigned — filing the friction, never the run report.