Ecom weekly trading deck

Weekly Trading Deck

Operating question: What happened this week, why, and what should the operator do next?

Run this as an evidence-first operating play. Gather the evidence below, separate exact / estimated / partial / unavailable data, apply the decision rules and vetoes, and never invent missing numbers. Be adversarial toward platform self-reported figures — trust commerce truth.

Evidence to gather

• Commerce (Shopify / Woo / BigCommerce, etc.) — this week vs last week: gross and net revenue, order count, AOV, units per order, refunds, and new vs returning customer split. This is your source of truth for revenue; GA4 is not.
• GA4 — for the matching window: sessions, engaged sessions, ecommerce conversion rate, and purchases, broken out by default channel grouping (Paid Search, Paid Social, Organic Search, Direct, Email, Organic Social, Referral). This is what lets you split a revenue move into the sessions × CVR × AOV levers.
• A comparison baseline — last week (WoW) and the same week last year (YoY) or your weekly target/forecast. A WoW move with no baseline is uninterpretable.
• Top movers — products and collections with the biggest revenue gain and biggest revenue loss vs last week.

Optional, if available:

• Promo / discount calendar — which days ran an offer, the depth, and the code. Without it you will misread a promo spike as demand.
• Site/deploy log — theme changes, checkout edits, app installs/removals, payment-method changes timestamped to the day.
• Stock position on hero SKUs — a revenue drop on a top product is often a stockout, not a demand or CVR problem.
• Device split (mobile vs desktop CVR) — the single fastest way to localize a checkout regression.
• Tracking health — known consent-banner, GTM, or pixel incidents that would dent measured CVR without denting real demand.

How to decide — in order

1. Establish the revenue move and its size. Net revenue this week vs last week, as a number and a percent, commerce-side. Note the WoW and the baseline (YoY or target) gap. Until you know the magnitude you can't tell signal from noise (see Decision Rules).
2. Decompose before you explain. Hold revenue = sessions × CVR × AOV and find which term carries the move. Sessions flat, CVR down → a conversion problem. CVR flat, sessions down → a traffic problem. Both flat, revenue down → an AOV / mix problem (or refunds). Never narrate the cause before you've isolated the lever — this is the step everyone skips.
3. Localize the moved lever.
   • Traffic move → which channel? Split sessions by default channel grouping WoW. A paid drop is a budget/delivery question; an organic drop is rankings/seasonality; a direct/email drop is often a tracking or send-calendar artifact.
   • CVR move → which surface? Split CVR by device (mobile vs desktop) and by landing/entry behavior. A mobile-only CVR drop with desktop flat is a checkout or page regression until proven otherwise — run the tracking check before you accept it as real.
   • AOV move → units-per-order vs price/mix. Did a promo shift the basket, did a high-AOV hero go out of stock, or did discount depth rise?
4. Attribute the cohort. Overlay new vs returning. New-customer revenue down is usually acquisition (a traffic/channel story); returning down is usually lifecycle/email or a loyalty/repeat issue. They route to different owners.
5. Name the top movers. Which 3–5 products/collections drove the gain and which drove the loss. One stocked-out hero can explain an entire "CVR drop" that isn't one.
6. Apply the vetoes, then write the narrative. Strip promo distortion, confirm tracking, then compress to one paragraph and exactly three actions, each with an owner and a recheck date.

The prompt to run

You are my ecommerce trading analyst running the "Weekly Trading Deck" play.

PRE-FLIGHT: First list which required inputs I provided vs. missing. If commerce
truth for the week (net revenue/orders/AOV/new-vs-returning) plus a
confirmed-trustworthy CVR/sessions read is missing, STOP and return only (a) what's
missing and (b) how to get it — never estimate it or proceed. Do not narrate a CVR
drop on unverified tracking.

GOAL: tell me what happened to revenue this week, isolate WHICH LEVER moved it
(sessions vs conversion rate vs AOV), attribute it to a channel and to new-vs-returning,
and hand me the THREE highest-priority actions with owners. Not a data dump.

I will paste: commerce revenue/orders/AOV/new-vs-returning this week vs last week;
GA4 sessions, ecommerce CVR, and purchases by default channel grouping (and by device);
top product/collection movers; my promo calendar; any deploys or stock events. Some
data may be missing.

RULES:
- Revenue = sessions x conversion rate x AOV. Always decompose the move into these three
  levers BEFORE explaining a cause. Name the single lever that carries most of the move.
- Treat commerce as revenue truth; treat GA4 as the sessions/CVR lens. Do NOT expect them
  to reconcile exactly and do NOT use GA4 revenue as the headline.
- Judge signal vs noise by traffic volume: on a store doing only a few hundred orders a
  week, a single-digit-percent WoW move is usually noise unless a lever shifted hard.
- A mobile-only CVR drop is a checkout/page-regression suspect: demand a tracking check
  before declaring a real CVR decline.
- Separate promo-driven spikes from underlying demand using my promo calendar.
- Split new vs returning to route the issue to the right owner (acquisition vs lifecycle).
- Every claim must carry: a number, source, time window, and confidence level.
- Separate exact / estimated / partial / unavailable evidence. Do not invent missing data.

RETURN:
1. A 3-4 sentence executive trading read in plain operator language.
2. The decomposition table, using exactly this header row:
   | Metric | This wk | Last wk | WoW | vs Target/YoY | Read |
   Use "—" for any cell you cannot fill from the evidence. Do not add or drop
   columns, and do not replace the table with prose.
3. Lever verdict: TRAFFIC, CVR, or AOV problem (or "noise"), with the channel/cohort.
4. Top movers up and down (product/collection, revenue delta, why).
5. Exactly 3 prioritized actions, each with owner, timing, and what would change the call.
6. Vetoes/caveats that downgraded confidence, and what evidence is blocked.

Decision rules

• KEEP — the number is inside its normal band: revenue WoW within roughly ±5% on a higher-volume store (1,000+ orders/wk) or ±10–15% on a lower-volume store, all three levers stable, no risk signal. Report it, don't act on it.
• WATCH — a lever moved but the read is early or directional: one soft week, a sample too small to be sure, or a window polluted by promo/stock. Set a recheck date and a threshold that would escalate it.
• REFRESH — a lever is decaying with a credible fix: CVR slipping on a specific surface, a fatiguing channel, a collection losing share — where the asset still has a viable path back.
• FIX — the read itself is unsafe: GA4 and commerce diverge beyond explainable drift, a tracking/consent incident is live, or refunds/test orders are polluting revenue. Fix measurement before you trade on it.
• KILL — stop or pull something only when the downside is clear, the sample is large enough to be sure, and no veto protects it (e.g. a money-losing promo mechanic that's tanking AOV with no offsetting volume).
• Every recommendation must include a number, source, time window, and confidence level — "CVR 1.8% (GA4, last 7d, high confidence)", never "conversion felt low."

Vetoes — stop if any apply

• Do not explain noise as signal on a low-traffic store. Under ~200–300 orders/week, a single-digit-percent WoW wobble is usually variance, not a trend. Name it noise and move on.
• Do not declare a CVR drop real until tracking is cleared. A consent-banner change, a GTM/pixel break, or a duplicated purchase event can fake a CVR move with zero change in actual demand. Run the tracking check first.
• Do not read a promo spike as underlying demand. A discount week inflates orders and AOV-down, and borrows demand from the following week. Separate promo days from clean days before drawing a trend.
• Do not blame "the site" when a hero SKU was out of stock. A revenue drop concentrated in one product is a supply problem, not a conversion problem — check stock before touching the funnel.
• Do not call causality from one metric. A channel-grouping shift, a calendar artifact, and a real demand change can all look identical at the headline. Triangulate with the decomposition and the cohort split.
• Do not recommend writes, budget shifts, refunds, customer messages, or catalog changes without an explicit human approval step.

Output

A short trading narrative (the paragraph the operator reads aloud), a decomposition table, the lever verdict, top movers, and exactly three prioritized actions with owners.

Decomposition table (minimum columns):

Metric	This wk	Last wk	WoW	vs Target/YoY	Read
Net revenue	$X	$Y	±%	±%	one-line
Sessions
Conversion rate
AOV

Action packet (minimum columns):

Action	Lever	Status	Evidence	Owner	Timing
Example	CVR	FIX	Source + number + window	Ecommerce + Dev	Today

---

Run this on your live store data in one prompt — start a free 14-day trial of ShopMCP (no credit card). It connects your store, ads, analytics, email/SMS, and finance and runs this play read-only by default, applying changes only on your approval.