Mercado Libre Market Intelligence: Category and Seller Data for LATAM Brands

Introduction

A brand analyst is asked for "our price position on Mercado Libre in LATAM." They pull a dataset, build a chart, and present an average price index across the region.

The number is meaningless. It averaged Brazilian reais against Argentine pesos against Mexican pesos, over a period in which one of those currencies moved substantially against the others. It also averaged five distinct catalogues with different assortments, so the products underneath the index were not the same products in each market.

Mercado Libre is not one marketplace. It is a set of national marketplaces sharing a platform. Getting LATAM intelligence right starts with accepting that and building five collections. Our Mercado Libre scraper service runs exactly this way — five country pipelines with a shared product identity layer rather than one regional job.

Five countries, five catalogues

The major country sites each operate their own catalogue, currency, seller base and category tree.

Practical consequences of the split:

• A product ID in one country does not resolve in another. There is no single cross-country product key. You have to build one.
• The category tree differs by country. Depth, naming and leaf structure are not identical, so a category mapping table per country is mandatory.
• A brand's official presence may exist in three countries and not the other two.
• Number formatting differs. Decimal comma versus decimal point, thousands separators, and currency symbols that look similar but are not. Parse to a numeric type with an explicit locale, never with a naive string strip.

Cross-country comparison therefore requires an internal product identity layer. You match listings across the five sites on brand, model or part number where present, normalised title tokens, pack size, and image similarity. That match is a probabilistic join with a confidence score, and it should be stored as such — not silently asserted.

Country-specific collection is described in our web scraping services in Brazil and web scraping services in Mexico pages.

Portuguese and Spanish are not interchangeable

Brazil is Portuguese. Everything else in the set above is Spanish. That affects more than translation.

• Keyword and search-term sets have to be built natively per language. A translated Spanish keyword list applied to Brazil will miss how Brazilian buyers actually search.
• Attribute values differ in vocabulary even where the concept is identical. Colour, material and size terms need per-language normalisation dictionaries.
• Regional Spanish varies. Mexican, Argentine, Chilean and Colombian product vocabulary is not uniform. Assuming one Spanish dictionary works across four markets produces silent matching failures.
• Review and Q&A text needs language detection per record rather than per country, because listings do contain mixed-language content.
• Store a listing_language field detected from the text, not inherited from the site. It catches the cases that break sentiment and keyword pipelines later.

Currency, inflation and the Argentina problem

Argentina is the case that forces a discipline everyone should have anyway.

In a high-inflation environment, a nominal price series in local currency tells you almost nothing about competitive position. A product whose ARS price rose 40% may have become cheaper in real terms. A flat nominal price is a substantial real-terms price cut.

Rules that make a LATAM price series usable:

• Store the observed price in the original currency, always. Never store only a converted value.
• Store the currency code explicitly on every row. Do not infer it from the country field. Listings can be denominated unusually.
• Store the FX rate used for any conversion as its own field, with its own timestamp and source. This is what lets you separate a supplier price change from a currency move.
• Timestamp at observation, not at load. A price series in a fast-moving currency needs observation-time precision, and a collection that ran over eight hours should not carry one date for all rows.
• Report a price index within country, and compare indexes across countries. Do not compare converted absolute prices across countries as a headline. An index normalises out both FX and general inflation.
• Never blend currencies into an average. If a stakeholder asks for one LATAM price number, give them a weighted index with the weighting method stated.

The same discipline protects the other four markets when their currencies move. Argentina just makes the consequence of skipping it immediate. The handling described here is implemented in our multi-currency price monitoring and cross-border price intelligence services.

Official store versus third-party seller

Mercado Libre distinguishes official brand stores from general third-party sellers, and the distinction drives most brand-monitoring value.

Why it matters concretely:

• Price integrity. A brand's own official store price is its intended position. Third-party listings of the same product show what the market actually charges, and that gap is the brand's real exposure.
• Authorised versus unauthorised. A brand monitoring distribution wants to know which sellers are listing its products and whether they appear on its authorised list. The seller identity is the whole answer.
• Counterfeit and grey-market signals. Anomalously low prices from unknown sellers, or listings with mismatched pack descriptions, are the detectable patterns. You find them by comparing third-party listings against the official store's own catalogue and price.
• Content control. Third-party listings use their own titles, images and descriptions. Where those diverge from brand assets, that is a measurable content-compliance issue.
• Buy-box-equivalent competition. Where multiple sellers offer the same catalogue product, which one is surfaced by default is a competitive outcome the brand should be tracking.

Capture a seller-type flag plus the seller identifier on every listing row. Maintain your authorised-seller list as reference data and join against it. The output is a per-country list of sellers with listing counts and observed price ranges — count of third-party sellers listing brand SKUs by country is the shape of the finding, and it should come from the data rather than an estimate. The seller-side extraction is described in our seller and vendor data service.

Shipping is part of the observed price

A LATAM price comparison that ignores shipping is wrong in a direction that varies by seller.

MercadoEnvíos and the platform's shipping programmes mean a listing's delivered cost can differ substantially from its listed price, and free-shipping eligibility is often tied to price thresholds and seller status. Two listings at the same price are not the same offer if one ships free and one does not.

What to capture:

• Listed price, separately from any shipping cost
• Whether free shipping is offered, as a boolean
• Shipping cost where shown, with currency
• Shipping programme or fulfilment indicator where exposed
• Any full/fulfilment-by-platform flag
• Estimated delivery window where shown
• The destination used for the query, because shipping cost is destination-dependent

That last field is the one teams forget. Shipping data is only comparable if you queried from a consistent destination. Pin a reference postcode per country and record it on every row.

What to collect: field schema

Traversing the category tree

Assortment questions need category coverage, not keyword coverage. Keyword collection finds what you thought to ask for. Category traversal finds what exists.

A workable approach:

• Walk the category tree per country from the root, recording node IDs, names, parent relationships and depth.
• Store the tree with a snapshot date. Trees change, and an assortment comparison across two dates needs to know whether a node moved.
• Identify leaf nodes and collect listings at leaf level. Mid-tree collection produces overlapping, hard-to-deduplicate result sets.
• Handle result-set depth limits. Category listings are paginated and typically capped. Where a leaf exceeds the cap, subdivide by price band, brand filter or sort order to enumerate the remainder.
• Record listings-per-node counts as a first-class metric. Category depth by country is one of the most useful outputs of the whole exercise, and it falls out of the traversal for free.

Cross-country category comparison then runs on your mapped internal tree, not the platform trees, which do not align.

The questions this actually answers

Worth being concrete about the output, because the collection effort should be justified by decisions.

Note that "share of listings" and "share of search" appear here and "market share" does not. Listings and search positions are observable. Market share is not, from this data, and claiming it would be an estimate dressed as a measurement.

Collection is limited to public-facing listing and category pages. Platform terms and applicable law differ across the five countries and change over time, so treat the compliance position as a per-market review rather than a single regional assumption.

Actowiz Solutions runs Mercado Libre collection as five country pipelines with a shared product identity layer, explicit currency handling and official-store versus third-party seller classification.

We report within-country indexes and observed seller counts rather than blended regional averages, and we hand over the FX and timestamp fields behind every figure.

You can also reach us for all your mobile app scraping, data collection, web scraping, and instant data scraper service requirements!

FAQ

Can Mercado Libre data from different countries be combined into one dataset?

It can be stored in one dataset, but only with country, currency and observation timestamp on every row, plus an internal product key that joins matched listings across sites with a confidence score. Country catalogues, category trees and listing IDs are separate, so there is no native cross-country product identifier.

How should Argentine peso prices be handled in a price series?

Store the observed ARS price and the currency code as recorded, keep any FX rate used as a separate field with its own timestamp and source, and report a within-country price index rather than a converted absolute price. In a high-inflation market a flat nominal price is a real-terms price cut, which only an index makes visible.

Why does the official-store versus third-party seller distinction matter?

The official store price is the brand's intended position, while third-party listings show what the market actually charges. The gap between them, and the identity of the sellers involved, is what brand teams use to find unauthorised distribution, price erosion and content that diverges from brand assets.