Manual Data Entry in Customs: Where Errors Begin

Every ocean import declaration in Brazil (the DUIMP, Brazil’s single import declaration) requires dozens of fields that already exist, with the same value, in the documents that accompany the cargo: gross weight appears in the commercial invoice, the packing list, and the bill of lading; the FOB value appears in the invoice and the freight quote; the container number appears in the B/L, the packing list, and the booking confirmation. The customs broker reads each value from the source document and types it into Siscomex, Brazil’s foreign trade system, field by field, shipment by shipment. Manual data entry is how most errors reach the DUIMP, because every re-keyed field is a chance for transposition, omission, or misinterpretation. Under the new per-field penalty regime, the cost of that re-keying has become measurable.

Source document data reaches Siscomex through re-keying

A typical ocean import in Brazil involves at least four source documents: the commercial invoice, the packing list, the B/L (or HBL), and the certificate of origin. Each document comes from a different party. The exporter issues the invoice. The freight forwarder or carrier issues the B/L. The chamber of commerce in the origin country issues the certificate. The forwarder’s operator in Brazil receives these documents, usually by email as PDFs, and types the data into the forwarder’s ERP. The customs broker then receives the same documents, or receives the data already entered by the operator, and types it again into Siscomex to register the DUIMP, the CE Mercante (Brazil’s cargo manifest system), and, when applicable, the LPCO (import license).

The same gross weight gets typed at least three times: by the operator into the ERP, by the customs broker into Siscomex, and, if there is an LPCO, by the analyst into the licensing module. If the B/L reads 12,500 kg and the operator types 12,050 kg because of a digit transposition, the customs broker who uses the ERP data as a reference may copy the wrong number into the DUIMP without knowing. If the broker checks the B/L directly and types 12,500 kg, the ERP data stays divergent, and the freight invoice, calculated by the carrier based on the B/L weight, will not match the forwarder’s system. Re-keying multiplies versions of the same data point, and each version can diverge for different reasons.

The number of fields a customs broker fills manually in a single DUIMP varies by cargo type and number of line items (additions). A DUIMP with five additions can have over 80 fields sourced from original documents. When the shipment involves an LPCO, a certificate of origin for a preferential trade agreement, and an invoice with multiple HS codes, the manually keyed fields can exceed 120 per shipment.

Re-keying means interpreting, and interpreting means risk

Manual data entry for trade documents goes beyond copying characters from a PDF into a form. The customs broker interprets what they read. When the commercial invoice says “stainless steel fittings, grade 304, assorted sizes” and the DUIMP field requires a goods description in Portuguese, the broker translates, classifies, and adapts. If the invoice description is generic (“machine parts”) and the HS code demands specificity (“stainless steel screws, 6 mm diameter”), the broker adds detail that the source document does not provide. Every addition is a judgment call, and every judgment call carries responsibility.

The same applies to units of measure. Invoices from American exporters list weight in pounds and dimensions in inches. Siscomex requires kilograms and centimeters. The conversion seems trivial, but across shipments with dozens of line items and mixed volumes, a conversion factor applied to the wrong value slips through visual review. A weight of 2,200 lbs converted with a factor of 2.2 instead of 2.205 creates a small per-item difference that, summed across 40 packages, diverges from the B/L gross weight.

Foreign currency values present the same challenge. The commercial invoice may list a unit price with four decimal places (USD 0.0875 per piece), while Siscomex requires the total item value with two. The rounding, multiplied by the number of pieces, can produce a cents-level difference between the invoice total and the sum of the line items in the DUIMP. Brazil’s Federal Revenue Service cross-checks the declared DUIMP value against the supporting documents, and any discrepancy, even in cents, can trigger a query during customs risk assessment.

The per-field penalty turns every keystroke into standalone risk

The former penalty regime under Article 711 of Brazil’s Customs Regulation (Decree 6,759/2009) assessed 1% of the customs value for inaccurate information, with a floor of BRL 500 and a ceiling of 10% of the total declaration value. A DUIMP with three incorrect fields on a USD 50,000 import generated a proportionally modest fine. The risk was diluted by the size of the shipment.

Supplementary Law 214/2025, regulated by Decree 12,955/2026, changed the unit of assessment. Article 341-G, item XIX, sets the penalty at 100 UPF per inaccurate, incomplete, or omitted piece of information, with a floor of 50 UPF and a ceiling of 1% of the total operation value per fiscal document. With the UPF set at BRL 200 in 2026, each incorrect field can cost up to BRL 20,000. Three incorrect fields in the same DUIMP generate three standalone infractions, according to analysis published by Conjur. The punitive enforcement of the new regime is scheduled for January 1, 2027; during 2026, the Federal Revenue Service has adopted a guidance-oriented posture. The law, however, is already in force.

For a customs broker who manually keys 80 fields per DUIMP from PDF documents, each field is a standalone bet. A transposed digit in the weight, a cent-level difference in the value, a wrong port code, an HS code with the last digit swapped: each one is, under the new regime, a separate infraction. When the broker processes 15 to 20 DUIMPs per day, the number of manually keyed fields can exceed a thousand, and so can the number of penalty possibilities.

Manual review breaks down at volume

Every experienced customs broker reviews what they typed. The standard practice is to open the source document on one screen and the Siscomex form on another, comparing field by field. This review works when the professional is rested, when the volume allows attention to each shipment, and when the source document is legible and well organized.

In practice, at a customs brokerage office in Santos or Itajai, a broker processes between 10 and 25 shipments per day during peak periods. If each shipment requires review of 80 fields, that is 800 to 2,000 visual comparisons per day. Research on human error in transcription tasks shows that the error rate for manual keying of numerical data falls between 0.5% and 1% under controlled conditions, with trained and rested operators. Under time pressure and high volume, the rate rises. With a thousand fields keyed per day, even a 0.5% rate means five fields with error potential. Five incorrect fields, under the new regime, can represent up to BRL 100,000 in accumulated penalties.

Cross-document checking, which we covered in detail in a separate post, helps catch discrepancies between the invoice and the B/L, between the packing list and the DUIMP. The problem is that manual cross-checking also depends on attention and time. The broker who has already reviewed 500 fields by mid-afternoon applies the same cross-check with less rigor on the last shipments of the day. The error that visual review should have caught escapes precisely when volume exceeds the professional’s attention capacity.

Validation at extraction eliminates data entry as an error source

When data is extracted automatically from the document, read by a trained model that identifies each field with its position on the page and a confidence score, the manual data entry step disappears. The container number is extracted directly from the document and validated against the check digit (ISO 6346) before it reaches the system. The destination port code is matched against the UN/LOCODE table. The currency code is checked against ISO 4217. The gross weight extracted from the B/L is compared with the invoice weight and the packing list weight, and any discrepancy is flagged before anyone opens the DUIMP.

This validation at the source changes the customs broker’s work. Instead of keying data and then reviewing it, the professional receives data already extracted and already checked against consistency rules. What remains for the broker is the part that requires human judgment: tariff classification, interpreting a certificate of origin with ambiguous wording, deciding which Incoterm to declare when the invoice is inconsistent. Manual data entry gives way to qualified review, which is the work the customs broker was trained to do.

The field that remains uncertain after automatic extraction goes to human review. The field that passed all validations continues as a clean record. In manual processing, every field depends on the same professional’s attention. With automatic extraction and validation, only the fields the system flagged as uncertain need human attention. The cognitive load drops, and the error rate in the final output drops with it.

Frequently Asked Questions

How many fields does a customs broker manually fill in a single DUIMP?

The count depends on the cargo type and the number of line items (additions). An ocean import DUIMP with five additions can have over 80 fields sourced from original documents. With an LPCO (import license) and a certificate of origin, the total can exceed 120 fields per shipment.

Does rectifying the DUIMP eliminate the penalty for inaccurate information?

Rectification corrects the field in the declaration, but the penalty for inaccurate information is triggered at the time of registration. The importer has 60 days to rectify after a fiscal notice, but rectification does not eliminate the applicable penalties, according to the Federal Revenue Service manual.

What is the difference between cross-document checking and validation at extraction?

Cross-document checking compares fields across different documents (invoice against B/L, for example) and can be done manually or with automation. Validation at extraction happens earlier: when data is read from the document, it is checked against structural rules (check digits, ISO codes, value ranges) before it enters the system. The two complement each other.

How Autodocs validates fields before anyone opens the document

Autodocs classifies the incoming document, extracts each field with its value, confidence score, and source page, and applies its validator library: container number by ISO 6346, air waybill by IATA mod-7, port by UN/LOCODE, currency by ISO 4217, plus rules the customs broker or freight forwarder defines. A second model re-reads only the flagged fields. What remains uncertain goes to a person for review in the dashboard. What passes validation continues as a record ready for the destination system.

For the customs broker, the practical difference is receiving invoice, B/L, and packing list data already extracted and already checked against rules that would take minutes to apply manually to each document. The fields reach Siscomex from validated records, not from manual keystrokes.

Learn about Autodocs or get in touch.

Every field that moves out of manual data entry reduces the risk of rectification and frees the broker to focus on what demands judgment: interpreting the regulation, classifying the goods, and defending the declaration. A useful starting point is to count how many fields your team keys per day and calculate the accumulated risk under the new penalty regime.