DutyDelta — Tariff Change Impact

Tariff updates. A clearer view of your catalog.

Compare your product list against selected official HTS editions. See changed source information, unchanged items and codes that need another look. Add purchase values to focus attention where more of your supplied commitments are involved.

DutyDelta compares source information. It does not choose classifications or calculate your duties.

Try the demo · Compare your catalog · Pricing · Help

What you get

A before-and-after review and a CSV working list. Open REPORT.html for explanations and REVIEW.csv for the items to investigate. Missing source coverage remains visible rather than being labelled unchanged.

Start with the catalog alone. Purchase-line analysis and a review agenda are optional; neither is needed for a first comparison.

Demo

Leave Demo selected to explore three fictional items: changed, unchanged and unresolved. Start the run and open REPORT.html from Output. Demo rates are illustrative, not real applicable duties.

The demo has no custom item-review charge. Apify's automatic $0.00005 per start still applies; check the displayed maximum cost.

Start here

1. Select Compare. The bundled defaults compare 2026-18 with 2026-19.
2. Paste your catalog into Catalog CSV, using the template below.
3. Run and open REPORT.html. Download REVIEW.csv for your working list.

itemId,sku,htsCode,originCountry
0001,SAMPLE-A,2908.91.00.00,US
0002,SAMPLE-B,0101.21.00.10,CA

These are fictional products paired with official source captures. Supply your existing eight- or ten-digit HTS codes and keep IDs as text. The Actor does not guess classifications.

Add purchase values when useful

Link each purchase line to a catalog itemId:

purchaseLineId,itemId,declaredCustomsValueUsd,plannedEntryDate
po-1,0001,5000.00,2026-10-15
po-2,0002,,

The supplied USD value is the whole line total; quantity does not multiply it again. Blank values remain missing. Purchase values help organize questions, not estimate duties or savings.

Coverage and capacity

Bundled 2026-18 and 2026-19 editions support source-text comparisons for 26,779 codes. Another 27 indexed codes have unresolved source hierarchy. Unfound codes and unavailable evidence remain unresolved, not unchanged.

These are fixed editions, not today's latest tariff. Included coverage supplies the relevant source information; you do not need to prepare reference packs for it. A supported comparison does not determine the duty applicable to a shipment.

Up to 5,000 catalog rows and 10,000 linked purchase lines can be submitted, subject to input, source-evidence and result limits. Many distinct codes can reach a source-work limit before the row ceiling. Not every 5,000-code catalog fits one run.

Start small. If asked to split a catalog, use smaller groups, keep each purchase line with its matching item and include it only once. No additional run starts automatically; each run has its own pricing and limits. A source-evidence-limit refusal stops before catalog results and completed-item charges are submitted. Startup charges may still apply.

Pricing

$4.00 per 1,000 completed catalog-item reviews ($0.004 each at the base rate), including this Actor's platform usage, plus $0.00005 per start. No Dataset-row surcharge applies.

Apify plan	Price per completed item review
Free	$0.0040
Bronze	$0.0036
Silver	$0.0034
Gold / Platinum / Diamond	$0.0032

Completed eligible reviews use the catalog item as the billing unit; purchase lines, agenda questions and scenarios add no separate units. Required source gaps do not become completed reviews. Results are saved before the custom charge. Check current pricing and the displayed maximum before starting. Account subscriptions, taxes and unrelated services are separate.

An unknown charge may already have settled. Inspect the run rather than assuming a retry is free.

Focus your next review hour

The optional review agenda groups exact source-change questions shared across your catalog and links the affected purchase lines. Supply your effort estimates and up to two hour budgets to see a suggested order of work.

A planned question is not an answered question, and an answered source question does not clear an item for import. Known purchase values are commitments, not duties or savings. Missing estimates and amounts remain visible.

For example, two products can share one source-text question linked to $1,500 of known purchase value and another purchase with no value supplied. A half-hour estimate fits a half-hour budget but not a quarter-hour budget. The question is planned once, while each product's other requirements remain unresolved. A removed marker in source text does not by itself establish a changed duty rate.

The agenda settings below describe the optional setup. This feature is available in Compare, not Monitor.

Help

Ask about preparing a catalog, coverage or results in the Issues tab. Include a run ID for a run problem. Use fictional examples rather than private purchase files, credentials or access links.

A stopped run may already have results or charges. Inspect its Output and billing state before starting another. The Actor does not connect to your purchasing system or approve an import.

Advanced reference

Input and source limits

catalogCsv fields are itemId,sku,htsCode,originCountry; origin is a two-letter country code. Optional purchasesCsv links through exact item IDs. Duplicate/conflicting identities and invalid joins fail early.

For source editions, latest is rejected: select the explicit bundled editions or supply qualified prepared baseline/target packs. This source-edition rule is separate from the Actor's build tag. Supplied packs retain their own combined-input limits and provenance checks.

Bundled references load in bounded parts. Required source checks finish before comparison results are published. Missing or damaged parts are unavailable, not unchanged. Internal limits are 96 MiB of decoded reference work across both editions, 16 MiB of retained source evidence and 16 MiB of results. These are work/output limits, not predictions of memory use or duty amounts.

Review-agenda settings

In Compare, set reviewAgenda to {} to list questions in REPORT.html and OUTPUT.summary.reviewAgenda. Each question retains its exact edition, capture, source-node locator and bounded before/after code, description and rate-text fields. Other or nested changes require checking the cited source. Similar wording does not merge questions. Missing source coverage creates an obtain-evidence question.

Copy returned taskId values into reviewAgenda.taskEstimates with decimal-string effortHours and optional integer priority; add up to two decimal-string budgetsHours values. Stale, unknown or duplicate task IDs are refused. Hours range from 0 to 168, with up to six fractional digits; priorities range 0–1,000, default 100.

Lower priority numbers come first. Within a priority, explicit zero-hour tasks precede tasks ordered by greatest marginal known USD purchase value per supplied hour, then task ID. Overlap is recalculated after each selection; tasks that do not fit are skipped. This greedy method is transparent, not a guaranteed optimum. Missing effort remains unscheduled and missing value unknown. Zero effort and zero value are valid declarations.

Each scenario starts fresh and counts a literal purchaseLineId once even when it relates to several questions. Do not add overlapping question values as separate commitments. Every item retains all required question IDs and planned/unplanned IDs in the report and CSV. No identified question means only that this bounded comparison found none, not approval to import.

The optional agenda permits 512 tasks, 40,000 item/purchase links and 2 MiB of agenda JSON, within existing overall limits. Split work or omit the optional agenda when a bound is exceeded.

Output and repeated comparisons

REPORT.html is the review and REVIEW.csv the working list. OUTPUT links canonical JSON and files; Dataset rows are a convenience view. A failed Dataset append does not erase saved reports.

Monitor uses a named history store and state.scopeId, with optional baselineSnapshotId. CHANGES.json/CSV separates source, decision, clock and input changes. History is accepted for 90 days; missing or ambiguous history is disclosed. Select historyStoreId for hosted Monitor; any supplied state.storeId must match. This grants the stated read/write access, not a schedule. Agenda planning is not a Monitor feature.

Interpretation, privacy and recovery

Displayed rate text is source evidence, not an applicable-rate determination. Source changes need not change duties owed. Classification advice, legal eligibility, numerical duties, total-duty and savings calculations are not supplied. Unresolved coverage remains separate from unchanged items; known-value totals exclude missing amounts.

Use permitted business data and appropriate Apify access and retention settings. Limited permissions apply, with explicit grants for selected history storage. Keep credentials, unrelated personal identifiers and payment information out of input. Results retain selected evidence, not raw upstream error bodies or unrelated fields. The Actor does not automatically delete storage.

Publication and history intent are recorded before submission. An interrupted operation may be pending or unknown, not cancelled. An unknown total custom charge is null, distinct from an acknowledged event count. Re-entering the same interrupted run does not resubmit results or charges. Inspect existing output and billing before a new review; there is no cross-storage transaction guarantee. The charge event remains catalog-item-reviewed.

Existing saved workflows

New users can start with Demo. Numbered build 0.5.2, verified October 8, 2026, supports this guide's wider bundled coverage, interruption handling and optional review agenda. Use that exact build for a fixed selection; movable tags can change. Preserve original files and read the public release notes before changing an established task.

For existing older contracts, the October 3 inventory recorded 0.3.1 for wider bundled source coverage and 0.2.4 for the earlier comparison/Monitor contract. Those selectors predate the review agenda and later recovery correction. They are historical compatibility references, not a promise of current or indefinite availability. Inspect the actual selector and keep an existing pin until you choose a migration. Older pinned examples do not demonstrate later features.