Home / Guides / CBP Rulings

How Often Does CBP Revoke Its Own Rulings? A 2,350-Ruling Reliability Census

Importers lean on CBP binding rulings for one reason: they are written down, citable, and enforceable. The uncomfortable question every compliance team eventually asks is how much those promises decay — how often does CBP come back and revoke or modify its own classification calls? We answered it the direct way: a full census of the 2,350-ruling extraction of CBP's CROSS database carried on this site, counting every ruling flagged as revoked (35), every ruling flagged as modified (31), and every explicit revocation or modification edge between documents (142 and 102 respectively).

The short version: 1.5% of published rulings end up revoked, another 1.3% carry a modification flag, and when changes do happen they skew late— the median ruling that gets revoked survives roughly 6.4 years first. Low churn is good news for anyone treating a ruling as planning certainty; the tail is why monitoring still matters.

What we counted, exactly

The corpus is our local mirror of CBP's CROSS API (rulings.cbp.gov/api/): 341 search terms, each contributing up to a 20-result slice, deduplicated by ruling number, yielding 2,350 unique rulings issued between 1988-05-18 and 2026-09-23 (38 years) and retrieved on 2026-10-12. Each record carries structured flags plus the raw letter text; US-government-works licensing lets us republish derived counts freely.

• Revoked: the record's boolean is_revoked flag is true — 35 rulings, or 1.5% of the corpus.
• Modified: the record lists at least one entry in modified_by — 31 rulings.
• Edges: explicit cross-references where one ruling names another in a revokes or modifies field— 142 revocation edges and 102 modification edges, carried by 189 "carrier" rulings.

Two counting disclosures matter. First, node counts and edge counts disagree on purpose: an edge can point at a document outside our sliced corpus, so 142 edges coexist with only 35 records flagged revoked in-corpus. Second, the slice design means change counterparties are under-sampled: we could resolve the exact revoking document inside the corpus for only 10 of the flagged revocations (and 13 modifications). Survival-time statistics below use only those resolved pairs — a conservative, fully re-derivable subset.

The reliability census

Read as a reliability figure: a randomly drawn CBP ruling from the last 38 years carries a 1.5% chance of having been formally revoked and a 1.3% chance of carrying a modification reference, for a combined 2.8% touched-or-revoked rate. Notably, the two populations are disjoint in this corpus — no ruling is both revoked and separately modified — so the union is simply their sum, 66 documents.

Category mix is stark: every one of the 35 revoked rulings is a Classification ruling, which is unsurprising given that Classification dominates the corpus (2,333 of 2,350 records), but it means "binding ruling risk" is really "classification call risk". Origin and other categories show zero revocations in this window.

When the survivors were issued

The decade view exposes a recency caveat you should keep in mind when quoting the headline rate: rulings issued in the 2020s (227 of them) have simply not had enough calendar time to accumulate revocations, and the 1980s cohort is tiny. The 1990s cohort— 21 revocations against 854 issuances— is the deepest observed layer; its within-cohort revoked share of 2.5% is the closest thing to a mature-cohort estimate in the data.

How long does a ruling last?

For the 10 revocations whose revoking document also sits in the corpus, the ruling-to-revocation gap runs from 0.8 to 27.3 years with a median near 6.4; modifications resolve similarly (median 6.6 years across 13 resolvable pairs). Change activity is not a museum piece either: 13 of the 23 resolvable change events are dated 2016 or later.

Top 5 of 10 resolvable revocations by years survived; the full change log lives in CROSS.

The extreme case is worth a pause. Ruling 088585 (subject: "Thomas Skid Steer Loader; Works Trucks Fitted With Lifting And Handling Equipment; Loading; Unloading; Self Propelled Shovel Loaders", issued 1991-05-02) was revoked by H296917 on 2018-08-27 — after 27.3 years on the books. Any importer who treated the original letter as permanent infrastructure spent most of three decades exposed to that reversal.

Revocation versus modification in practice

Legally the two flags lead different lives. A revocation withdraws the ruling's binding effect going forward, which usually follows an adverse court decision or a CBP initiative against a whole product family; entries already made under the old letter can face a retroactive bill. A modification keeps the ruling alive but narrows or re-points its scope— the original remains citable, just not for the facts CBP changed. Both are proposed in advance in the Customs Bulletin with a comment window, so a weekly scan of the Bulletin for your headings converts a silent hazard into a scheduled chore.

Where changes concentrate

Counted once per changed ruling per cited heading; top 6 of 50 distinct headings shown.

Changed rulings cluster in exactly the product families where margin points are thinnest: tires (4011 leads with 14 changed rulings citing it), bed linens, luggage and seats. This is consistent with how revocation campaigns actually form — one reclassified product family anchors a series of follow-on reversals, which is also why the edges outnumber the flagged nodes.

What a prudent importer does with this

• Treat rulings as dated instruments. A classification letter older than a decade has survived the median survival gap of revoked rulings by definition; pull the current CROSS status before you cite it in an entry summary, not after a CBP audit does.
• Watch the Customs Bulletin, not just CROSS. Revocations and modifications are proposed publicly before they take effect; the notice-and-comment step in 19 CFR 177.12 is your early-warning channel, and it exists precisely because retroactive surprises are costly.
• Prioritize monitoring by exposure. With a base revoked share of 1.5% and clustering in tires, textiles and seats, a monitoring budget aimed at your actual HS families beats blanket anxiety. Our ruling search surfaces the change flags alongside each decision.
• Price the downside, not the average. The distribution has a fat tail: a small number of rulings survive decades while others flip within a year. Quote-margin decisions should key off your family's history, not the corpus mean.

Frequently asked questions

1. How often are CBP rulings revoked?

In our 2,350-ruling census of the CROSS database (issued 1988-05-18 through 2026-09-23), 35 rulings carry the revoked flag — 1.5%. Another 31 rulings (1.3%) are flagged as modified rather than revoked.

2. What is the difference between revocation and modification?

A revocation cancels the ruling's binding effect outright; a modification amends its scope while leaving it in force. In the corpus the two never overlap on the same ruling: 35 revocations plus 31 modifications give 66 uniquely changed documents out of 2,350.

3. How long does a typical CBP ruling stay valid?

Among the 10 revocations whose revoking document is also in the corpus, survival ranged from 0.8 to 27.3 years with a median near 6.4 years. Modifications show a similar median of 6.6 years.

4. Can I rely on an old CBP ruling for classification?

Yes, with hygiene. Age alone is not disqualifying — most rulings are never revoked — but the older the letter, the more likely its product family has seen related reversals. Check the current status on CROSS, look for successor rulings that cite the original (our reversal tracker catalogs 142 such references), and monitor the Customs Bulletin for proposals affecting your headings.

5. Where can I check whether a specific ruling is still in force?

The official CROSS database shows each ruling's current status. ExportSonar's ruling search layers the revocation and modification flags onto the same records so you can screen a shipment's ruling without leaving the tariff lookup workflow.

Where to check next

• most disputed HS codes at CBP
• classification ruling reversal pairs
• CBP ruling search
• duty calculator