How Cofactr Uses AI + Big Data to Demystify Tariffs

Are you getting invoices with tariff charges you didn’t expect? Tariffs look like a line item, but they behave more like a layered system with missing inputs. At Cofactr we evaluated the three specific mysteries that make tariffs difficult to predict, then used AI and real transaction data to remove the veil.

You might assume this is straightforward. Look up the tariff rate, apply it, move on. That’s not how it works in practice.

Right now, tariffs in electronics sourcing are shaped by overlapping policies. Section 301 tariffs apply to goods from China. Section 122 tariffs apply selectively to certain HTS codes. That means you need two inputs before you can even begin to estimate a rate.

• Country of origin
• HTS classification

Neither is as clean as it sounds.

Country of origin is especially slippery. A single component can be fabricated in one country, packaged in another, and distributed globally through multiple channels. The official country of origin depends on specific rules about substantial transformation, and those rules are not always intuitive.

It gets more complicated. The same exact part number can legitimately have multiple possible countries of origin depending on the specific lot you receive. One shipment might qualify as originating in Malaysia, another in China, based on where final transformation occurred. You don’t know which one applies until the distributor actually fulfills your order and declares it.

So even when you think you’ve identified the country of origin, what you really have is a set of possibilities, not a confirmed answer.

So who determines country of origin? The manufacturer does. More precisely, they declare it. You rarely get documentation backing that claim, and you almost never get visibility into how that determination was made.

HTS classification has its own issues. Components that look nearly identical on a BOM can fall under different codes depending on subtle technical distinctions. A minor classification difference can change the tariff exposure significantly.

So before you even get to pricing, you’re already dealing with inputs that are partially hidden and occasionally ambiguous.

Even if you had a perfect tariff rate, you still wouldn’t have the number you care about.

Tariffs are assessed on the importer’s declared value at the border. That’s the distributor’s cost, not the price you pay.

This distinction matters more than most teams expect.

The distributor’s cost includes what they paid the manufacturer. That number is tightly held. It directly reflects their margin structure, and distributors protect that information aggressively.

What you see is resale pricing. That’s already marked up. The tariff, however, was calculated on a lower, hidden number.

So when you try to back into tariff impact from your purchase price, you’re working from the wrong baseline. You’re effectively estimating a percentage of a number you don’t have.

Let’s say you somehow knew the tariff rate and the distributor’s cost. You still wouldn’t know what shows up on your invoice.

Distributors decide how to incorporate tariffs into their pricing. There’s no single standard.

Some treat tariffs as part of cost basis, then apply their normal margin on top. That’s consistent with traditional distribution economics. Others may pass through tariffs more directly, or adjust pricing dynamically based on market conditions and inventory position.

In practice, this leads to wide variation.

• Two distributors can source the same component
• Pay similar tariffs at the border
• Charge meaningfully different tariff line items to the customer

That variation isn’t necessarily arbitrary. It reflects differences in sourcing strategy, inventory timing, and pricing models. But from the buyer’s perspective, it feels unpredictable.

At this point, you’re dealing with three layers of uncertainty. Rate, cost basis, and pass-through behavior. Stack them together and you get what most teams experience today, inconsistent and hard-to-explain tariff charges.

Tariffs don’t just affect unit price. They influence sourcing decisions, supplier selection, and even design choices when teams start considering alternate parts or geographies.

If your tariff assumptions are off, you can easily mis-rank suppliers or underestimate total landed cost. That shows up later as margin compression or budget overruns.

And because tariffs are often treated as a line item after the fact, they don’t always get the same scrutiny as component pricing. That’s where surprises creep in.

We approach the tariff problem from three angles: classification, real-world transaction data, and duty recovery.

AILANA: Resolving Country of Origin and HTS Classification

We built an AI system called AILANA. One of its jobs is to determine country of origin and HTS code for individual components at scale.

Instead of relying solely on distributor declarations, AILANA analyzes part data, manufacturing information, and supply chain signals to generate a best-fit classification. Is it perfect on every line item? No. But it moves the starting point from guesswork to a structured, repeatable process.

Learning From Actual Transactions

Classification gets you to a rate. Transaction data gets you to reality.

We operate as a procurement platform, which means we see a large volume of real transactions — not quoted tariffs, not theoretical rates, but the actual tariff charges that appear on invoices. Over time, patterns emerge. Some suppliers consistently pass through close to the estimated tariff. Others apply higher effective charges, often because tariffs are folded into their margin structure.

We quantify this behavior using what we call a tariff pass-through ratio. A ratio near 100% suggests tariffs are passed through roughly as incurred. Above 100% indicates additional markup layered on top. Lower ratios can appear in competitive situations or when inventory was sourced under different conditions.

We apply that pass-through ratio to the tariff assessed based on AILANA's classification. The result is what we call the Estimated Supplier Tariff Surcharge — a forward-looking estimate of what will actually show up on your invoice. Not theoretical modeling. Derived from observed outcomes across many real transactions.

Recovering What You've Already Paid

Predicting tariff exposure is only half the equation. If you import components and export finished products, you may be eligible to recover up to 99% of the duties you've already paid — a program called duty drawback that's been part of U.S. trade law since 1789.

To help customers act on that, Cofactr has partnered with Forge, a Y Combinator and Google-backed company that handles the entire drawback filing process. Because Cofactr already manages your procurement and import data, most of the legwork is done from day one. Forge identifies eligible refunds, files with U.S. Customs and Border Protection, and only takes a percentage when refunds are received — no upfront cost, no operational changes required.

U.S. Customs estimates over $6 billion in eligible drawback refunds go unclaimed every year. For hardware manufacturers already navigating margin pressure from tariffs, that's worth a look.

When you combine AILANA's classification with transaction-based pass-through data, you get something more useful than a static tariff rate. You get an expectation of what will actually be charged.

That distinction matters.

Instead of asking, "What is the tariff on this part?" you start asking, "What will this supplier likely charge me for tariffs on this part?" Those are not the same question.

Our position is that these estimates are the most accurate available at a market level — not because every single line item is exact, but because the model reflects real supplier behavior over time.

And once you know what you're likely to pay, the next question is what you can get back. Between forward-looking tariff estimates and duty drawback recovery through Forge, the goal is the same: shrink the gap between what tariffs cost you on paper and what they actually cost your business.

For a hardware team, this changes how you plan and buy.

• Fewer surprises when invoices arrive
• More accurate landed cost calculations during quoting
• Better comparisons across distributors

It also gives procurement teams leverage. When you understand how a supplier tends to apply tariffs, you can factor that into negotiations or sourcing strategy.

Tariffs aren’t going away, and they’re not getting simpler. Policy changes, supply chain shifts, and geopolitical factors will keep introducing variability.

What changes is your ability to navigate that variability.

When tariff behavior is treated as unknowable, teams default to buffers and assumptions. That adds cost and reduces confidence in planning.

When tariff behavior is measured and modeled using real data, it becomes another variable you can manage.

That’s the shift Cofactr is aiming for. Not eliminating complexity, but making it observable and predictable enough to act on.

Tariffs feel unpredictable because key inputs are hidden, distributor costs, classification decisions, and pricing strategies. Once you break those pieces apart, the problem becomes clearer.

Our approach combines AI-driven classification with large-scale transaction data to estimate what actually matters, the tariff charges you’ll see in real purchasing scenarios.

That doesn’t make tariffs simple, but it does make them legible.