Methodology
Every formula, every source, and every known limitation. If you are going to plan spending against these numbers, you should be able to check them.
The formulas, in full
There is nothing proprietary here and nothing withheld for the paid tier. These formulas are industry standard; the work is in applying them to costs people forget.
Break-even ROAS = price ÷ (price − landed − platform fee − payment fee − shipping − returns)
Break-even CPA = AOV × contribution margin
Target ROAS = 1 ÷ (contribution margin − target net margin)
MER = total revenue ÷ total ad spend
Landed cost = FOB + freight + duty + packing + clearance
Duty = customs value × combined rate
Two details that trip people up. Customs value is the goods plus buyer-paid packing invoiced separately — not landed cost, and not FOB alone. And returns are modelled as price × return rate × loss factor, where the loss factor is how much of the sale you fail to recover; set it to 1 for a full refund with no resale.
Where the tariff data comes from
Last verified 2026-08-22. MFN rates are pulled from the USITC Harmonized Tariff Schedule REST interface (2026 Revision 16), not transcribed from a secondary source.
We model duty as stacked layers rather than a single number, because that is what it is:
- MFN general rate — from the tariff schedule, per HTS code. Source: USITC HTS
- Section 301, China lists — 7.5% (List 4A) or 25% (Lists 1–3), per HTS code, China only. Unaffected by the February 2026 Supreme Court ruling. Source: USTR
- Section 301, forced labor — effective 12:01am ET 24 July 2026. China 12.5%, Vietnam 12.5%, India 10%. Source: USTR Docket 2026-0265/0266
The reason for the layered model is practical: when one layer changes — and three changed during 2026 — you need to know which, and so do we, to tell you what moved.
What these estimates get wrong
This is the section most tools do not write. Ours has four entries and we would rather you read them before trusting a number.
- Categories are not HTS codes. Duty applies per 10-digit code. Our ten categories are approximations built from a representative code, and the spread inside a category can be enormous — knit apparel ranges from 2.6% to 32%. Every result shows the representative code and the full range.
- We model percentage duties only. Some headings carry a specific duty as well — cents per kilogram, cents per pair. Footwear, woven apparel and bags all contain such codes. Where they apply, our figure is too low, and the result says so.
- Exemptions are disclosed, not modelled. Annex I of the forced labor notice exempts civil aircraft parts, pharmaceutical-use articles, Section 232 goods (steel, aluminium, copper, vehicles) and charitable donations. Nothing in our ten consumer categories is exempt, but if you import metal drinkware or aluminium cases, check the Section 232 carve-out yourself.
- Platform fees change without notice. TikTok Shop’s US referral fee moved during 2026 and public sources disagree about where it landed. Rather than pick one and present it as fact, we make it an input — your Seller Center is authoritative for your account.
Everything here is a planning estimate. Your customs broker’s classification determines what you actually pay, and no calculator changes that.
Category coverage
Ten categories, each with the HTS headings it covers, the representative rate we calculate with, and the real range across that heading.
| Category | HTS | Used | Range |
|---|---|---|---|
| Apparel | 6109–6110 | 16.5% | 2.6–32% |
| Apparel+ specific duties | 6203–6204 | 27.9% | 0–27.9% |
| Footwear+ specific duties | 6402–6404 | 20% | 0–48% |
| Bags, luggage & wallets+ specific duties | 4202 | 17.6% | 0–20% |
| Consumer electronics | 8517–8518 | 0% | 0–0% |
| Cables, chargers & electrical accessories | 8544 | 2.6% | 0–5.3% |
| Home & kitchen | 3923–3926 | 3.4% | 0–6.5% |
| Home textiles | 6302–6304 | 11.3% | 0–20.9% |
| Toys & games | 9503 | 0% | 0–0% |
| Jewelry & accessories | 7117 | 11% | 3.9–11% |
Benchmark data
Last reviewed 2026-08-22. We show channel ROAS figures as a spread rather than a single median, because the published sources do not measure the same thing — some quote medians, some means, some loose ranges — and for the same channel they differ by more than 2×.
Presenting them as one authoritative number would be tidier and less true. The disagreement is itself the useful finding: no benchmark knows your cost structure, which is why the calculators derive your floor instead of comparing you to an average.
How often this is updated
When a rate changes, within 24 hours, and the verification date on every page moves with it. We do not refresh dates on a schedule to look active — a date here means someone checked the source that day.
Rate changes also get a changelog entry, so you can see what moved and when rather than finding your numbers quietly different.
Who makes this
A one-person operation, built because the existing tools each solve half the problem. Tariff calculators tell you what you owe but nothing about whether the product still works. Profit trackers tell you what happened after you spent the money. Neither tells you what return you need before you commit.
No claim to being a customs authority — the sourcing discipline on this page is the substitute for credentials. Everything traces to a primary document, and where it cannot, that is stated rather than papered over. Corrections are welcome, particularly on classification.