Dynamic Pricing Ecommerce

Dynamic Pricing for Ecommerce

Turn seller-approved economics and trusted signals into a bounded repricing system with explicit rules, approvals, monitoring, and a kill switch.

Installation

npx skills add nexscope-ai/eCommerce-Skills --skill dynamic-pricing-ecommerce -g

Capabilities

• Define SKU eligibility for automatic, approval-required, or manual repricing.
• Calculate contribution-safe floors and commercially justified ceilings.
• Select demand, inventory, competitor, season, and promotion signals without treating noisy observations as facts.
• Create deterministic rule matrices with bounded price steps, cooldowns, and conflict precedence.
• Simulate normal, downside, promotion-stack, stockout, and price-war scenarios.
• Design approval, audit-log, rollback, anomaly-breaker, and emergency-stop controls.
• Produce a staged platform implementation and measurement plan without enabling live changes.

Usage Examples

Design safe Amazon repricing rules for these 200 SKUs without starting a price war.

Create an inventory-aware dynamic pricing plan for my Shopify store.

Which products can be auto-repriced, and which should always require approval?

Audit these existing repricing rules for margin, promotion, and rollback risks.

Inputs and Collection

Use seller-supplied and inspected evidence first. Collect:

• SKU, variant, channel, marketplace, currency, tax treatment, fulfillment method, and lifecycle stage;
• current price, realized selling price, list or compare-at price, coupons, promotions, bundles, and discount-combination rules;
• COGS, inbound freight, duties, packaging, fulfillment, payment, referral, affiliate, ad, return, and other variable costs;
• target contribution dollars or margin, approved floor, approved ceiling, and brand or MAP constraints;
• inventory on hand, inbound stock, sell-through, age, weeks of cover, replenishment lead time, and stockout risk;
• timestamped traffic, orders, units, realized price, conversion where available, cancellations, and returns;
• comparable competitor offers with variant, pack size, seller, fulfillment, availability, delivered price, source, and capture time;
• current repricing tool, platform capabilities, rule cadence, account permissions, approvers, and business objective.

If required economics or authorization details are missing, ask one consolidated follow-up. If they remain unavailable, design a provisional system but mark affected floors, rules, and automation decisions as blocked.

Workflow

1. Establish the Evidence Boundary

List the exports, pages, cost sheets, platform settings, and seller facts actually inspected. Label each material input:

• Confirmed: supported by inspected evidence.
• Assumption: an explicit scenario placeholder, not an observed fact.
• Unknown: missing information that blocks reliable automation.

Do not invent demand, competitor history, costs, fees, elasticity, conversion, or platform capability. A visible competitor price is a point-in-time observation, not a durable market signal.

2. Calculate Economic Guardrails

Use realized seller-funded economics:

Net Revenue = Selling Price - Seller-Funded Discounts - Refund Allowance
Contribution $ = Net Revenue - COGS - Variable Selling Costs
Contribution % = Contribution $ / Net Revenue

When percentage fees apply to selling price:

Price Floor = (Unit Cost + Fixed Variable Costs + Target Contribution $) / (1 - Variable Fee Rate)

Model base, high-return, high-ad-cost, promotion-stack, and fee-change cases. Keep a contractual or legal minimum separate from the calculated economic floor. Define a ceiling from value, reference-price, policy, and customer-trust constraints; do not create artificial scarcity or an inflated reference price.

3. Classify SKU Automation Eligibility

Assign each SKU to one control tier:

Default uncertain SKUs to the more restrictive tier. Automation convenience is not evidence that a SKU is safe to automate.

4. Select and Validate Signals

For every signal, record source, freshness, coverage, failure mode, and fallback:

• Competitor: only normalized, comparable, available offers; reject mismatched packs, used items, suspicious sellers, and stale captures.
• Demand: use observed seller traffic and orders with timestamps; separate price effects from ads, content, seasonality, and stock.
• Inventory: use on-hand, age, sell-through, lead time, and replenishment risk; do not treat a feed error as surplus or scarcity.
• Time or event: use scheduled windows with explicit start, end, timezone, and promotion interaction.
• Own promotion: distinguish seller-funded from platform-funded incentives and confirm whether discounts stack.

Never use protected personal characteristics or opaque customer vulnerability to set individualized prices. Avoid price-gouging, collusion, and discriminatory outcomes.

5. Build the Rule Matrix

Each rule must specify:

Use deterministic rules first when data is sparse or explainability matters. An algorithmic recommendation still requires the same economics, input-quality, authorization, and rollback gates.

6. Simulate Before Enabling

Replay or model at least:

• ordinary demand and competitor movement;
• a competitor stockout or feed disappearance;
• an extreme competitor price or mismatched offer;
• promotion and coupon stacking;
• a high-return or fee-change downside;
• low inventory, excess inventory, and replenishment delay;
• repeated undercutting that could create a price loop;
• stale or unavailable input data.

Report rule firings, resulting price, contribution, approval path, clipped actions, and stop conditions. If reliable historical data is unavailable, use clearly labeled synthetic boundary cases rather than pretending to backtest.