.svg)
.svg){kind=icon}
Ask an operations team why a shipment went red and you will usually hear a version of the same answer: bad luck. The container ahead of it cleared, this one did not, and nobody can explain the difference. Luck is a comfortable model because it closes the conversation. It is also the wrong model.
The decision is made by a system that Article 43 of the Ley Aduanera places at the center of every clearance. It runs on data that was transmitted before the truck reached the booth. The authority does not publish the weight it assigns to each piece of that data, and any account claiming to know those weights is guessing. What is published, however, is enough to replace luck with something operationally useful: a risk model with identifiable inputs, some of which an importer controls and some of which it does not.
We have operated at the US and Mexico border for 122+ years. We clear more than 190,000 customs operations per year across 39+ ports with a 99.8% accuracy rate, and we hold three Patentes Nacionales in Mexico (Monterrey, Nogales and Manzanillo) alongside a US corporate customs brokerage license. That volume gives us a clear view of what correlates with selection, and of where the public record stops. This piece stays on the right side of that line.
The mechanism, as the law defines it
Article 43 sets a strict sequence. Once the pedimento (Mexico's customs declaration) is prepared and the duties and, where applicable, trade remedy duties are paid, the goods are presented to the customs authority and the automated selection system is activated. That system determines whether a reconocimiento aduanero applies. If it does, the authority performs the examination in front of whoever presents the goods in the recinto fiscal.
The law also defines what that examination is. Article 2, fraction XV describes the reconocimiento aduanero as the examination of the goods and their samples that authorities carry out to gather elements confirming the truth of what was declared, and compliance with the provisions that tax and regulate entry and exit from national territory. The examination tests the declaration, not the shipment in isolation.
That is the fact most operations teams underweight. The file precedes the container. By the time a driver reaches the module, every variable the system can read has already been transmitted and paid.
One more feature of the current design matters. Mexico eliminated the segundo reconocimiento, the second examination performed by authorized private inspectors, through the reform published in the DOF on December 9, 2013. The legislative record for that reform stated the objective plainly: move from a system built on reviewing large volumes quickly toward a customs operation that performs fewer reviews with a high rate of effectiveness in detecting irregularities. Every design decision since has followed that direction.
There is also a window that closes. Under Article 89, taxpayers may amend the data contained in the pedimento before activating the automated selection system, except in the cases that require prior authorization from the authority. After activation, an amendment is no longer a correction. It is a finding.
What is published about the inputs, and what is not
Start with scale, because scale explains the design. In a June 2025 government forum presentation on customs intelligence, the project team working on the risk model reported ANAM figures of nearly 22 million cargo operations per year, of which 1.6 million resulted in reconocimientos, requiring more than 12,000 hours daily. The same document states the constraint directly: there is neither the operational capacity nor the infrastructure to review 100% of operations.
Those two figures put the ratio at roughly seven examinations for every hundred cargo operations. The arithmetic is ours, not the agency's, and the underlying counts cover cargo operations rather than pedimentos alone, so treat it as an order of magnitude rather than a rate you can plan against.
The same presentation names the variables the risk model uses to classify an operation. They fall into two groups. The first describes the transaction: tariff classification down to chapter, heading, subheading and NICO, the port of entry, the hour at which the automated selection is activated, the payment method, the country of origin, and the mode of transport. The second describes the party: the importer's RFC, the declared type and value of the goods, the prior tax compliance history, and comparative behavior patterns among similar importers.
Read that second group again. Your declarations are not evaluated only against the law. They are evaluated against your own history and against companies that look like you.
The design intent is equally explicit. The model is described as producing decisions that are deterministic yet unpredictable, executing without human intervention, with risk parameters updated automatically and an audit trail preserved. The stated goal is that a red light should carry a high probability of predicting an actual irregularity, without slowing clearance to the point of stopping commerce.
What is not published, and what we will not invent
The weight assigned to each variable. The threshold that converts a risk score into a red light. Inspection rates by port of entry, by sector or by tariff chapter. The cadence at which parameters recalibrate. None of it appears in the Ley Aduanera, in its Reglamento, in the Reglas Generales de Comercio Exterior or in any published ANAM statistic we can verify. Any table that assigns a percentage to each "cause" of a reconocimiento is inference presented as fact. It is also worth noting that the presentation above describes a project and its design criteria, not a published operating manual for the system running at a given port today.
The inputs you control
Everything on the transaction side of that list is produced by your own file. The 2026 framework has made the contents of that file explicit rather than customary.
Article 59, fraction V now requires the electronic file to contain the information and documentation proving the resources used to carry out the operation. The decree lists them: the guarantee referred to in Article 36-A, fraction I, paragraph e); the CFDIs; commercial invoices or equivalent documents; electronic payment transfers or letters of credit; transport, insurance and related service costs; contracts related to the transaction; the documentation supporting the concepts added to or excluded from transaction value under Articles 65 and 66; and any other document established through rules that demonstrates the operation actually took place. That is no longer a recommended practice. It is a list, published in the DOF on November 19, 2025.
Four disciplines follow from it.
Documentary coherence. The value on the invoice, the payment that settled it, the freight and insurance charged, and the value declared on the pedimento have to tell one consistent story. Subvaluation is named in the government presentation as one of the core problems the model was built to detect, alongside fractioning, shell companies, inexact classification and contraband. A file that cannot reconcile its own value chain is the profile the model was designed to find.
Classification consistency over time. The same product should carry the same Mexican tariff code with NICO across every entry, every supplier and every port. Where genuine doubt exists, Article 47 allows importers, exporters and agencias aduanales to formally consult the SAT on classification when a product could fall under more than one code or when the applicable code is unknown, and Article 49 bis routes advance rulings under international treaties through the same authority. A documented consultation converts a recurring risk into a resolved one. It also matters after the fact: Article 199, fraction I reduces the fine by 66% when an omission results from inexact classification within the same heading and the description, nature and characteristics of the goods were correctly declared.
Inventory control that answers on demand. Article 59, fraction I requires automated and permanent inventory control systems that keep the control data for foreign trade goods updated at all times and available to the customs authority. For IMMEX operations, this is where temporary imports either reconcile or do not.
Correction before activation. The Article 89 window is the cheapest control in the entire process, and the one most often missed because it depends on catching an error in the hours before a truck is dispatched rather than the days after.
If your team has never audited its own pedimento file against the Article 59 list as published, that is the highest-yield review available this quarter. Our compliance team runs it as a discrete engagement for importers operating across multiple ports. Talk to a Joffroy trade compliance expert.
The inputs you do not control
The rest of the variable list is not yours. The port of entry is set by your logistics network. The hour of activation is set by your carrier's schedule and the queue ahead of it. The mode of transport, the country of origin and the tariff chapter your product falls into are determined by the product itself and by your sourcing strategy. The general direction described for the model is to move away from reviewing by tariff code and toward concentrating on the sectors, ports and transport modes carrying the highest risk, which means an entire category of goods can carry elevated exposure independently of how any single importer behaves.
Then there is capacity. The model design explicitly accounts for the availability of inspection personnel and for variable merchandise traffic. A port running short-staffed on a heavy day is a real input, and it is nobody's file.
The honest conclusion is the one the government presentation itself states: no matter how good an operator is, the possibility of being reviewed always exists, and what changes is the probability. That reframes the operational objective. The goal is not zero reconocimientos. Zero is not purchasable. The goal is that every reconocimiento you receive is short, documented and uneventful, because the file behind it answers every question the examination can raise.
What the 2026 reform changed here, and what it did not
The reform published in the DOF on November 19, 2025, in force since January 1, 2026, modified, added or repealed provisions across more than a hundred articles of the Ley Aduanera. Article 43 is not among them. The trigger, the activation point and the legal definition of the reconocimiento aduanero are the same text they were in 2025.
What changed is everything feeding the decision.
Article 6 now provides that customs authorities will enter into agreements with the Agencia de Transformación Digital y Telecomunicaciones (ATDT) for technology management and data analysis, and the second transitory article required those agreements to be signed within 180 calendar days of entry into force. Articles 10, 14, 14-A and 14-D now require recintos fiscalizados and authorized facilities to run technological systems integrating electronic inventory control, video surveillance, security, traceability and real-time monitoring, interoperable with the electronic customs system and with continuous remote access for the authorities. Article 144 closes with a new paragraph confirming that authorities may support their powers with systems, technological equipment, video recording equipment or any other available means.
Responsibility moved as well. Article 54 holds both the agente aduanal and the agencia aduanal responsible for the truth and accuracy of the data supplied, the correct determination of duties, the customs regime, the tariff classification and the NICO, and for confirming that the importer holds the documents proving compliance. Article 53, fraction II makes the agentes aduanales who are partners in an agencia aduanal jointly liable for the operations it clears, and Article 195 charges the resulting fines to them.
The Reglamento de la Ley Aduanera followed. Its reform decree was published in the DOF on February 23, 2026 and took effect the next day, completing the cycle the sixth transitory article of the November decree had ordered. We covered its operational consequences in our analysis of the new Reglamento.
The net effect is worth stating precisely, because it is easy to overstate. Nothing in the 2026 package changed what the automated selection system is or when it fires. What the package did was widen, deepen and accelerate the evidence base available to whatever model sits behind it.
What certification changes, and what it does not
This is where honest description matters most, because certification is frequently sold as a shortcut past the red light.
What the government actually publishes is narrower than that. SAT's own description of the Operador Económico Autorizado (OEA) modality of the Registro en el Esquema de Certificación de Empresas, grounded in Article 100-A of the Ley Aduanera and rules 7.1.1 and 7.1.4 of the Reglas Generales de Comercio Exterior, lists prioritization in the clearance of goods and the use of exclusive lanes for import and export. The published benefits catalogue adds concrete remediation rights: when excess or undeclared goods are detected during a reconocimiento aduanero or the exercise of verification powers, a certified importer has 10 days to file the corresponding permanent import or export entry and pay the fine under Article 185, fraction I, and 3 days for goods matching the registered production processes of an IMMEX program.
Read that catalogue closely and a pattern emerges. Certification changes how you are handled once selected, and what a finding costs you afterward. It does not appear, in any published rule we can cite, as a variable that lowers the probability of selection itself.
There is one place where your compliance record does carry an explicit legal consequence. Articles 98 and 99 govern revisión en origen for registered companies, and Article 99 computes each importer's margin of error from two populations that are defined precisely by selection: contributions declared on entries that were not subject to a reconocimiento aduanero, verification in transit or other review, against omissions detected on entries that were. The law does not just observe your error history. It calculates with it.
The 2026 reform also tightened entry to the scheme. Article 100-A added requirements that no partner has been convicted of crimes carrying a custodial sentence and that the company holds no administrative sanctions for the import or export of goods, while Article 100-C now bars a new registration outright where cancellation followed a final conviction for tax crimes or sanctions related to the entry and exit of goods.
So the defensible statement is this. Certification is a strong instrument for cost, predictability and remediation speed, and a public signal of a documented compliance system. Treating it as inspection insurance is not supported by anything the authority has published. We make the same distinction with clients weighing OEA, CTPAT and IVA/IEPS certification side by side, because the three solve different problems.
Where the file meets the container
Operators who have replaced the luck model with a risk model behave differently in one observable way. They stop treating the reconocimiento aduanero as an event that happens at the port and start treating it as an audit of work completed days earlier, in the file.
That is the posture we operate from. Our clearance teams reconcile the value chain before transmission rather than after a finding, lock tariff classification at the master data level so one code follows a product across every port and every entry, and treat the Article 89 window as a standing control rather than an emergency measure. Across 190,000+ operations a year, that is what produces a 99.8% accuracy rate, and what turns a red light from a disruption into a routine morning.
The published direction of travel points one way, from the 2013 decision to trade volume of reviews for precision through the 2026 obligations on electronic files, traceability and data analysis. The file you transmit is the version of your company the authority evaluates. Make it the accurate one.
TRADE. UNDER CONTROL.
