Americanas Data Scraping Services

The instalment plan is the headline, and it needs three fields

A Brazilian listing typically leads with an instalment presentation — a number of monthly payments and the amount of each — and shoppers compare on that figure at least as much as on the total.

• The instalment count matters on its own. A longer plan at the same total is a materially better offer to a shopper managing monthly outgoings.
• Whether the plan carries interest changes the total, and an interest-free plan is a genuine commercial concession rather than a presentation choice.
• The monthly amount is the number in the largest type on the page.

We capture instalment_count, instalment_monthly and instalment_interest_free as separate fields alongside the cash price, and never fold any of them into a single price column. A competitive comparison that ignores the plan misses the field the market actually competes on.

Payment method changes the price, so it is a dimension

Listings frequently show a lower price for particular payment methods — a discount for bank-slip or instant-transfer payment, against a card price that supports instalments.

This creates a genuine trade-off rather than a single best price: the lowest figure and the instalment option are usually mutually exclusive. A dataset that records the lowest number reports a price nobody choosing instalments can obtain.

We capture payment_method_prices as a set with each method's price, and record which figure the listing presented as primary. Both the lowest available price and the instalment-supporting price are retained, and any computed comparison states which basis it used.

Freight is per CEP, so a national price does not exist

Brazil is large and freight varies substantially by destination postcode. A single national delivered price is not a thing that exists.

We capture freight quotes against a stated CEP panel, with the seller's origin where displayed, and publish the panel alongside any national figure rather than quoting one destination as if it described the country.

On low-value items freight can exceed the item price, so a delivered comparison is frequently the only one that means anything — and it is always specific to a route.

State tax varies inside one country, so a price gap can be statutory

Brazilian consumption tax on goods is levied at state level, and the rate applying to a sale depends on the origin and destination states rather than on the country. That means a displayed price can differ between two Brazilian destinations for reasons that have nothing to do with a pricing decision.

• An intra-country price comparison can attribute to commercial strategy a gap that is entirely a matter of statute.
• The origin state matters, not just the destination, because the rate depends on the pair.
• Marketplace sellers sit in different states, so two offers on one product can carry different tax positions.

We capture the destination as a CEP and the seller origin where displayed, and record the displayed price gross with the tax basis as shown. We do not compute a net-of-tax figure ourselves — the rate for an origin-destination pair is a determination we cannot make reliably from public data, and asserting one would be inventing precision. What we supply is the gross price with both endpoints recorded, so a client with the rate tables can do the normalisation and check it.

Every marketplace seller has its own freight curve

Freight depends on the route, and on a marketplace each seller ships from its own origin. That means one product does not have a freight curve across the CEP panel — it has one per seller.

The consequence is that the cheapest seller changes by destination. A seller who is cheapest delivered to one region can be the most expensive delivered to another, and both statements are true simultaneously.

We capture freight per seller per CEP rather than per product per CEP, with the seller origin where displayed. Any delivered-price ranking is therefore specific to a destination, and where a national ranking is wanted it is computed across the published CEP panel with the panel stated.

Collapsing this to one freight figure per product is the most common way a Brazilian delivered-price analysis produces a ranking that is wrong for most of the country.

Marketplace sellers sit alongside own stock

Americanas sells its own stock and hosts marketplace sellers on the same product pages, and the two differ in what the price represents and in who owns the delivery promise.

We capture seller_type and the seller name where displayed, plus which offer held the default position. Where the page gives no indication the field is recorded as unknown rather than assumed to be the retailer.

For brand clients this is the channel-integrity field: marketplace sellers listing a brand's products outside its distribution arrangements are invisible in any dataset that stores one price per product.

Portuguese product data and Brazilian identifiers

Titles, attributes and category paths are in Portuguese, and titles frequently carry pack, variant and promotional information inline rather than in structured attributes.

We retain the published title and category path with translation supplied as separate fields, and publish parser accuracy against a labelled hold-out set rather than asserting it. EAN is retained where present, which in this market is inconsistent enough that attribute-based grouping is often needed as a fallback with the rule published.

Prices stay in real with currency on every record and no silent conversion; the daily rate is supplied separately where a converted view is required.