Marketplace Merchandising at Scale: Managing Billions in Online Retail

What Merchandising Means in a Marketplace

Merchandising is one of the most misunderstood functions in ecommerce. Many people think it's just picking products to sell. It's much more than that.

Traditional retail merchandising is about managing the customer in-store experience: what products are where, how they're displayed, what promotions drive traffic, how inventory flows through categories. It's about optimizing shelf space and customer journey.

Marketplace merchandising is different. You're managing not one store but multiple storefronts (often geographically distributed). You're managing not just your own products but third-party seller products. You're optimizing not just physical shelf space but digital space (search, category pages, recommendations, email campaigns, social media). And you're managing incentives: how do you motivate sellers to stock the right products, price competitively, and maintain quality?

At scale, marketplace merchandising involves:

• Range Selection: Which products should be available? How many SKUs per category?
• Inventory Management: Ensuring sufficient stock to meet demand without excess inventory tying up capital
• Pricing Strategy: Setting pricing to optimize revenue, margin, and competitiveness
• Promotional Planning: Running campaigns (weekly deals, flash sales, seasonal promotions) that drive traffic and conversion
• Content and UX: Product titles, descriptions, images, and page layout that drive conversion
• Seller Management: Educating sellers on what products to stock, how to price, how to maintain quality
• Data Analysis: Analyzing what's working (conversion, margin, velocity) and optimizing continuously
• Cross-Channel Coordination: Managing differences between online, mobile, and in-store experiences

Effective marketplace merchandising requires synchronization across all these dimensions. You can't optimize pricing without considering inventory. You can't run promotions without considering supplier capacity. You can't expand range without considering customer ability to discover products.

Managing 800+ Storefronts at Coles

At Coles, managing a $1.1B P&L across 800+ storefronts meant managing essentially 800 different customer bases with different preferences, different competitor landscapes, and different inventory availability.

A shopper in central Sydney had access to a different range, different competitors, and different delivery options than a shopper in regional Tasmania. Trying to merchandise the same products with the same strategy across all 800 locations would have been inefficient and poor customer experience.

Localized Range Selection: Rather than a single assortment across all stores, we implemented localized range management. Each store (or cluster of similar stores) had range decisions informed by local demand data. If a particular suburb had high demand for organic products, we'd stock more organic items. If another suburb preferred value brands, we'd adjust range accordingly.

This required data infrastructure: demand forecasting by store, by category, by brand. Then, merchandising decisions were data-driven rather than based on opinion or historical practice.

Store-Level Pricing Flexibility: Similarly, pricing varied by store based on local competition and demand elasticity. A suburb with strong competitor presence required competitive pricing. A suburb with limited competition could sustain higher margins. Manually setting prices for 800 stores wasn't feasible, so we implemented dynamic pricing rules informed by market data and margin targets.

Promotional Planning Across Locations: Centralized category managers planned promotions (weekly deals, seasonal campaigns), but execution was localized. A promotion might run across 200 stores in Sydney but a different promotion in Brisbane reflecting regional preferences. Automated workflows ensured consistency in core promotions while allowing localization where needed.

Automating with IT Teams: Managing 800 storefronts with entirely manual processes isn't possible. We built systems where:

• Demand forecasts automatically generated replenishment recommendations by store
• Pricing rules automatically set prices based on cost, demand, and margin targets
• Promotional calendars automatically executed across aligned stores
• Inventory alerts automatically flagged low stock or overstock situations

These systems didn't replace merchandisers—they freed them from manual work to focus on strategy: which categories to expand, which products to delete, how to respond to competitive threats, how to drive customer engagement.

Category Management: Each major category (produce, dairy, meat, pantry, etc.) had a category manager responsible for strategy. But rather than managing one category, they managed it across all 800 locations, adjusting range and promotions based on local insights. This created accountability and expertise depth.

Weekly Deals and Promotional Operations at Catch

At Catch, a $200M P&L marketplace on the Mirakl platform, promotions were the primary traffic and conversion driver. We ran weekly deals—every Monday we'd release a new set of 50-100 products at significant discounts for one week. This required complex orchestration across many functions.

The Weekly Cadence: Every Tuesday planning started for the following Monday's deals. This 10-day cycle involved:

• Merchandising Analysis (Tuesday): Review previous week's deals performance. Which products sold well? Which didn't? Update inventory levels. Identify slow-moving inventory that needs clearance.
• Range and Pricing (Wednesday): Propose next week's deals. Ideally deals are high-velocity products at 20-30% discount, driving volume and customer satisfaction. Avoid deals on slow movers—better to clear them with higher discounts or delete entirely. Negotiate deals with suppliers to achieve target margin (40%+).
• Content Creation (Wednesday-Thursday): Create marketing content. Deal hero images (lifestyle photography showing products in use), descriptions highlighting value proposition. A good deal image increased click-through by 15-25%.
• Email and Social Planning (Thursday): Plan email campaigns (send Tuesday night for Monday launch, Wednesday and Friday reminder emails). Plan social media posts (Instagram, Facebook, TikTok). For top 20 deals, we'd create social content Wednesday-Thursday for Friday/Saturday distribution.
• Warehouse Preparation (Thursday): Alert warehouse teams about incoming deal volume. Allocate space, prepare packing stations. High-demand deals required staging to avoid picking bottlenecks.
• Final Review (Friday): Review all creatives, confirm pricing, confirm inventory availability, run final checks on systems.
• Launch (Monday): Go live. Monitor real-time performance, be ready to pause deals if issues arise (e.g., out of stock faster than expected).

Cross-Functional Coordination: This weekly machine required tight coordination across:

• Merchandising: Deal selection, pricing, margin optimization
• Production/Studio: Creating hero images and product photography
• Content: Writing product descriptions and marketing copy
• Email and Social: Campaign planning and execution
• Warehouse: Space allocation, picking optimization, quality assurance
• Customer Service: Preparing for surge in inquiries (common with deals)
• Finance: Monitoring margin, discount impact, profitability per deal

Executing this weekly with consistency and quality requires a disciplined, experienced team and clear workflows. A failure in any step (wrong image, incorrect pricing, warehouse not ready) cascaded. We invested in people, processes, and systems to get it right.

Measuring Success: Each weekly deals cycle was measured:

• Number of deals launched and inventory sold through
• Revenue contribution to weekly total (typically 20-30%)
• Customer acquisition from deals (new customers buying for the first time)
• Repeat purchase rate (customers who became repeat buyers after a deal)
• Margin achieved (compared to target)
• Inventory velocity (average days to sell through deal)

We tracked this religiously and updated strategies based on data. If a category consistently underperformed on deals, we'd either change the approach or deprioritize. If a certain discount level consistently sold out in 2 days (suggesting price too low), we'd adjust.

Optimizing Gross Margin to 40%+ on Mirakl

At Catch, we achieved gross margin of 40%+ on the Mirakl platform—significantly higher than industry norm (25-30% for marketplaces). This didn't happen by accident. It was the result of deliberate merchandising and pricing strategy.

Margin Components: Gross margin comes from several sources:

• Supplier Pricing: We negotiated cost of goods with suppliers. Buying in volume, committing to steady offtake, paying on favorable terms—all created leverage for better pricing.
• Retail Pricing: We set customer-facing prices to maximize margin while remaining competitive and driving volume.
• Category Mix: Some categories have inherent higher margin (electronics, home & garden) than others (grocery, FMCG). By emphasizing higher-margin categories, overall margin improved.
• Customer Segment Mix: Different customer segments have different margin profiles. New customers often received discounts to acquire them; repeat customers paid higher prices. By tracking customer lifetime value, we optimized acquisition vs margin trade-off.
• Operational Efficiency: Fulfillment cost (picking, packing, delivery) impacts net margin. By optimizing fulfillment operations and automation, we reduced cost per order from 18% to 12% of revenue.

Pricing Discipline: Higher margin requires pricing discipline. We didn't discount indiscriminately. Every promotion was analyzed: does the volume uplift justify the margin reduction? A promotion that increased volume 40% but reduced margin 60% was a value-destructive deal, even if it moved product.

We tracked margin by product, by category, by promotion, by customer segment. We identified low-margin products and either increased price (if competition allowed) or discontinued them. We identified high-margin, high-velocity products and invested in marketing to expand distribution.

Supplier Partner Strategy: Margin optimization included supplier relationships. Rather than treating suppliers as vendors to squeeze, we partnered with top suppliers to build loyalty. We committed to volume, paid on time, gave them insights into what was selling. In return, they gave us better pricing, early access to new products, and co-marketing support.

Inventory Management: Days on Hand Under 45 Days

Inventory is either a competitive advantage or a cash drain. At Catch, we managed to keep inventory at under 45 days on hand (DOH), meaning we turned inventory roughly 8 times per year. This is very fast for a marketplace.

Fast inventory turn has benefits: lower capital requirements, fresher products, reduced obsolescence risk, more cash available for growth. But it requires precision.

Demand Forecasting: The foundation of fast inventory turn is accurate demand forecasting. We built forecasting models using historical sales data, seasonality, promotions, and market trends. For top 500 products (which represented 70% of volume), we had weekly demand forecasts.

Inventory Replenishment Rules: Rather than having buyers manually order inventory, we implemented automated replenishment rules: when inventory falls below minimum threshold, order replenishment quantity based on forecasted demand. Rules accounted for supplier lead times, demand variability, and safety stock.

Promotion Planning and Inventory: Promotions created demand spikes. We'd forecast deal demand, then align inventory purchases to support deals. A deal using 1,000 units required pre-positioning that inventory before the deal launch. Failing to do so meant lost sales; over-positioning meant excess inventory post-deal.

Slow-Moving Inventory Management: Inevitably, some products didn't sell as forecasted. We implemented reviews every 4 weeks: products with sales velocity below target were actioned—either clearance (heavy discount to move), bundle (combine slow mover with fast mover), or delete (stop selling). We were ruthless about this: no product stayed in inventory indefinitely without justification.

Category Balancing: Within categories, we managed assortment by velocity. High-velocity SKUs (best sellers) got more inventory depth. Low-velocity SKUs got minimal inventory. We'd delete the lowest-velocity 5% of SKUs quarterly, reallocating inventory to opportunities.

Dark Store Range and Merchandising

Dark stores (dedicated online fulfillment centers) require different merchandising approach than traditional retail stores.

Range Depth vs Breadth: A retail store might stock 50 SKUs of apples; a dark store stocks 5-8. The dark store range focuses on high-demand, high-velocity items. Specialty items and niche products are either excluded from dark stores or sent from a central warehouse.

Inventory Positioning: Dark store inventory is positioned for picking efficiency. Organize by category in linear flow, minimize picker movement. This is opposite to retail stores where shelf placement is optimized for customer journey and cross-selling.

Quality Management: Dark store staff are trained on quality standards—they must select optimal ripeness for produce, correct expiration dates for perishables, undamaged items. Quality audits are regular, and staff are incentivized on accuracy and customer satisfaction, not just speed.

Assortment Localization: Different dark stores in different areas serve different customer demographics. We'd adjust assortment by local preference: a dark store in an affluent area stocks more premium brands; one in a value-conscious area emphasizes value brands. Demand data guided these decisions.

Automating Workflows with IT and Data Teams

Merchandising at scale requires automation. Manual merchandising can't manage 800 storefronts or weekly promotions at scale with consistency.

Demand Forecasting Systems: We built statistical models (time series, regression, machine learning) that predicted demand by product, by location, by day. These systems updated daily, ingesting point-of-sale data, inventory levels, and market data.

Pricing Optimization: Rather than buying decisions pricing by category, we built systems that automatically set prices based on cost, demand elasticity, margin targets, and competitor pricing. A buyer might set a margin target (e.g., 35%) and competitor price band (e.g., within 5% of competitor), and the system automatically set prices within those constraints.

Promotional Automation: Promotional calendars were planned manually by category managers, but execution was automated. Once a promotion was approved, systems automatically:

• Updated product prices across all channels (website, mobile, store systems)
• Triggered email campaigns at scheduled times
• Updated search and category pages to highlight promoted products
• Alert warehouse teams to prepare for demand surge
• Monitored inventory levels and paused promotion if stock ran too low

Inventory Alerts and Workflow: Systems automatically identified situations requiring action:

• Overstock (inventory above target): alert buyer, consider clearance or discount
• Stockout: alert buyer, consider prioritized replenishment or temporary delisting
• Slow mover: alert category manager, consider discontinuation
• Promotion performance: alert merchandiser if deal underperforming

Each alert came with recommended action, but decisions were human. Automation provided visibility and suggestions, not replacement of judgment.

UX and Conversion: Achieving 18% Uplift

Merchandising isn't just about what products are available—it's about how customers discover and purchase them.

At Catch, we achieved an 18% conversion uplift through systematic UX and merchandising improvements:

Product Content Quality: Better product titles, descriptions, and images drove higher conversion. We invested in professional product photography (studio shoots showing products in context). We wrote detailed descriptions highlighting key attributes, benefits, and specifications. Product titles were optimized for search and clarity (vs. vague titles). These changes improved conversion by 6-8%.

Reviews and Social Proof: Customer reviews are powerful conversion drivers. We invested in post-purchase email campaigns requesting reviews. We displayed review ratings and counts prominently. We responded to negative reviews publicly, addressing concerns. Robust review systems improved conversion by 4-5%.

Search and Navigation: A customer who can't find what they want doesn't buy. We invested in search quality (using data to improve relevance), category navigation (clearer hierarchies), and faceted filtering (allowing customers to narrow results by attribute). Better search improved conversion by 3-4%.

Personalization: Different customers have different preferences. We invested in personalization algorithms that showed different products to different customers based on browsing history and purchase history. Personalization improved conversion by 2-3%.

Page Layout and CTA: Merchandising extends to page design. Where is the "Add to Cart" button? How prominent are product images? Where are reviews? How are related products positioned? Systematic A/B testing of page layouts improved conversion by 1-2%.

The 18% total uplift came from dozens of small improvements, each carefully measured and validated. This is the nature of high-performance ecommerce: small gains compound into significant results.

Platform Differences: Mirakl vs Marketplacer

Catch used Mirakl; Coles used a custom-built platform built on Marketplacer. Merchandise strategy differs between platforms because their business models and capabilities differ.

Mirakl Model: Mirakl is a multiseller marketplace platform. Catch leveraged Mirakl to enable multiple sellers, but also sold first-party inventory. Merchandising focused on curated selection (which sellers to partner with, which products to feature), seller enablement (training sellers on how to price, how to merchandise), and competitive positioning (where Catch competes vs. where sellers compete).

Marketplacer Model: Coles built on Marketplacer, a platform enabling marketplaces. Coles was the primary merchant, with minimal third-party sellers. Merchandising focused on range management, pricing optimization, and promotional execution across channels (online, mobile, in-store).

Key Differences in Approach:

• Seller Management: Mirakl requires seller recruitment, training, and performance management. Marketplacer relies on internal operations.
• Pricing Control: Mirakl has less pricing control (sellers set their own prices); Marketplacer has full control.
• Assortment: Mirakl assortment depends on seller offerings; Marketplacer assortment is fully planned internally.
• Margin Structure: Mirakl margin comes from commission on seller sales plus first-party margin; Marketplacer is primarily first-party margin.
• Fulfillment: Mirakl involves managing seller fulfillment quality; Marketplacer is primarily first-party fulfillment.

Both models can be successful, but merchandising approach differs significantly. Mirakl requires more focus on network effects and seller incentives; Marketplacer requires more focus on operational excellence and inventory management.

Conclusion: Merchandising as Science and Art

Effective marketplace merchandising at scale requires both art (judgment, creativity, intuition about customer preferences) and science (data analysis, statistical models, rigorous testing). The companies that win at scale are those that combine both.

You must be disciplined about what data reveals: inventory velocity, margin contribution, customer conversion patterns. But you also must apply judgment: understanding customer psychology, responding to market shifts faster than data lags suggest, making bold bets on new categories or products.

At Coles and Catch, our success came from building teams that combined merchants (experienced in product selection and customer psychology) with data scientists (building models and systems). Merchants made decisions, but those decisions were informed by data. Data systems identified opportunities, but humans decided how to act on them.

If you're building a marketplace, don't treat merchandising as afterthought. Invest in the right team, build the right systems, measure the right metrics, and optimize relentlessly. Your margins and customer satisfaction will reflect it.