Repricing—dynamically adjusting prices based on the competition—has become essential on Amazon, Cdiscount, eBay, and Mirakl. If configured incorrectly, it triggers destructive price wars that erode your margins. When configured correctly, it secures the Amazon Buy Box, drives sales, and protects your profitability. An in-depth look at a modern multi-marketplace repricing engine.

Why automate pricing?

On competitive marketplaces, prices change every 10–30 minutes among your competitors. Manually adjusting 2,000 SKUs across 6 channels is humanly impossible. Without automation:

You constantly lose the Amazon Buy Box (and thus 70–90% of your sales on those products)
Your prices remain static while the market shifts
You miss out on price-hike windows (competitor out of stock = opportunity to raise prices)
Your teams spend their days staring at Excel spreadsheets

The 3 approaches to repricing

1. Rule-based (business rules)

The most common and predictable approach. You define rules such as: “For this product group, I want to be €0.02 below the cheapest competitor, with a floor of 85% of the list price.” The engine recalculates and publishes.

Advantages: total control, explainability, no unexpected behavior. Disadvantages: you must update the rules as the competitive landscape changes.

2. Algorithmic (optimization)

The engine uses an optimization algorithm (gradient descent, bandits, reinforcement learning) that seeks the price that maximizes an objective function: revenue, margin, Buy Box share, unit sales. More powerful but less transparent. Best suited for mature teams and large catalogs.

3. Hybrid (most effective in practice)

We combine business safeguards (floor price, minimum margin, maximum price difference vs. competition) with more granular optimization within these safeguards. This is the approach taken by the Pixee PIM Repricing Engine.

Essential rules for any engine

The floor price

The most critical rule. Without a floor price, a competitor’s error (e.g., a price set to €1 by mistake) can trigger a domino effect. The floor price must be:

Calculated per product (not as an overall percentage)
Based on actual cost (purchase + fees + logistics) + absolute minimum margin
Non-negotiable: no rule above it can fall below this

The price ceiling

Less well-known but just as important. It prevents you from publishing an absurd price if your competitors are all out of stock and their “price” is artificially inflated by scrapers.

Recalculation frequency

A good engine recalculates at least every hour for high-turnover products, and every 10–15 minutes for Buy Box-critical SKUs. The frequency must be rule-based: recalculating every 5 minutes on all SKUs is unnecessary and costly in terms of API quota.

Outbound rate limiting

Amazon SP-API limits price updates to ~10,000 requests per hour. If your engine tries to recalculate 50,000 prices at once, you’ll be blocked. The engine must spread out the updates over time and prioritize the highest-stakes products.

Architecture of the Pixee PIM Repricing Engine

Pixee PIM v5.6 introduced a visual repricing engine. Its architecture:

Competitor scraping: scheduled jobs (hourly Celery Beat) that collect competitor prices via marketplace APIs (Amazon BuyBox, Mirakl offers, Cdiscount data) + smart scrapers for third-party sites.
Target price database: indexed Postgres table (marketplace, product, competitor, price, timestamp). Six months of history retained to detect patterns.
Visual rules: graphical interface where the user creates rules (IF margin > 20% AND competitor < X THEN my price = min(competitor - €0.02, RRP × 0.95)). No coding required.
Decision engine: for each product, applies rules by priority, calculates the new candidate price, checks safeguards, and saves the decision (with justification).
Deferred publication: updates are pushed to marketplaces via connectors, adhering to rate limiting and exponential retry.
Audit log: every price change is logged (old price, new price, triggered rule, reference competitor)—essential in the event of an Amazon dispute or internal audit.

Common mistakes that cost you

Ignoring shipping costs: You reprice based on the base price, whereas the Amazon Buy Box compares price plus shipping costs. Include them in the objective function.
No detection of competitor stockouts: if your main competitor is out of stock, you can raise your price. The engine must identify these situations and adapt its strategy.
A single rule for the entire catalog: your €10 products and €500 products don’t have the same price elasticity. Segment your rules by product family.
Repricing against low-quality competitors: following a seller with a 2-star rating and a 50% return rate makes no sense. Filter your benchmark competitors based on their feedback.
No "pause" mode: during marketing campaigns (Black Friday, sales), you need to be able to temporarily disable repricing without losing your configuration.

Observed ROI

On a catalog of 3,000 SKUs across 5 marketplaces:

+35% Amazon Buy Box win rate on average after 4 weeks of tuning
+12% sales volume without a drop in average cart value (hybrid repricing prevents price dumping)
Margin preserved: safeguards block rules that would fall below the target profitability
Pricing team time reduced by 6x — managers shift from a monitoring approach to a strategic management approach

How much does it cost?

A SaaS repricing engine typically costs between €100 and €500 per month, depending on the number of SKUs and marketplaces. Pixee PIM includes repricing in its mid-tier and enterprise plans. ROI is achieved within weeks in the vast majority of cases.

Conclusion

Automated repricing is not dumping: it is a pricing management discipline that protects your margins, optimizes your market share, and frees your teams from repetitive tasks. When properly configured, with robust safeguards and explicit decision rules, it becomes a sustainable competitive advantage. To discover the integrated Repricing Engine: Pixee PIM. For strategic guidance on your pricing: contact Pixee Play.

Automated multi-marketplace repricing: how it really works