Dynamic pricing and demand forecasting

Real company context

Price Shoes is a 100% Mexican company founded in 1996, a leader in catalog sales of footwear, apparel and accessories, present throughout Mexico and part of Latin America. It operates with over 35,000 products distributed across 100 different catalog versions, supported by a network of over 9,000 employees and roughly 900,000 partner-sellers who sell the products, plus 17 large-format stores and 22 community stores.

In 2025, Price Shoes launched a real digital-transformation project with Freeway (a Salesforce partner), focused on real-time personalization and AI, and on an intelligent loyalty program based on a recency-frequency-monetary (RFM) model for its partner-sellers. This project is real and verifiable, but its documented focus is personalization and loyalty — not dynamic pricing or demand forecasting.

The challenge (real business context)

Price Shoes' model has a distinct feature versus traditional ecommerce: catalog prices are set per campaign (multi-week cycles), not in real time like a marketplace. That means "dynamic pricing" can't apply the same way it would on a traditional marketplace — the real challenge is different:

• 35,000+ SKUs across 100 catalog versions: deciding price and quantity to produce/buy for each, campaign after campaign, with high risk of overstock or stockout.
• Online channel vs. physical catalog: the digital channel does allow more agile price adjustments than the printed catalog, but requires clear rules to avoid price conflicts between sellers offering the same product on different channels.
• Strong seasonality: back-to-school season, Buen Fin, Christmas, seasonal shifts (rain shoes, sandals) — demand for each category varies widely between campaigns.
• Decentralized selling: real demand depends on hundreds of thousands of partner-sellers selling across different regions, not a single centralized channel — making it hard to anticipate what product will run out where.

Application scenario (hypothetical, illustrative)

A demand-forecasting and pricing-support system, adapted to this campaign-based catalog model, could be structured like this:

DEMAND FORECASTING BY SKU AND CAMPAIGN

• Time-series model + exogenous variables (seasonality, comparable historical campaigns, category trends, regional weather) to predict how many units of each SKU will sell in the next campaign, by region.
• This informs early purchase/production decisions, reducing both overstock (tied-up capital) and stockouts (lost sales and seller frustration).

DIFFERENTIATED PRICING IN THE DIGITAL CHANNEL

• For the printed catalog, price stays fixed per campaign (a real business-model constraint).
• For the online channel, a pricing engine could adjust promotions and discounts within a pre-approved range, based on the product's real sales velocity during the campaign — without breaking the base-price parity that protects sellers.

SEASONAL-CATEGORY INVENTORY OPTIMIZATION

• Anticipating stockouts in high seasonal-turnover categories (sandals in spring, boots in winter) before they happen, adjusting purchasing and store/DC distribution by region.

Metrics this type of initiative typically aims to move

Illustrative values — not results reported by Price Shoes.

Metric	Typical goal of this type of project
Stockouts in high-turnover SKUs	Significant reduction
Excess inventory at campaign close	Reduced tied-up capital
Demand-forecast accuracy per SKU/campaign	Improvement over manual/simple historical forecast
Protected margin in digital channel	Promotion adjustment without eroding base catalog price