Why don't my DROP hashes match my records?
DROP sends each identifier as a SHA-256 hash in standard Base64, with no salt. You standardize your own values with the same rules, hash them, and compare the two strings exactly. There is no fuzzy match and no partial match: for NDZ and NameVIN only the full composite hash counts, so a right name and birth date with a wrong ZIP is still a miss.
A hash changes completely when one input character changes. So a slip in standardization does not produce an error message. It produces a hash that matches nobody, and the work item is reported as status 5 (not found) even though the person is in your data. The regulations ask brokers to apply any standardization they know will make a match more likely (11 CCR § 7613(a)(1)(A)).
The 12 slips below come from the rules in CalPrivacy's technical documentation, version 1.3.0 of 19 November 2026. Each example was computed by the code CalDrop Compliance uses to match batches; the hashes are cut to 12 characters to keep the blocks readable. The full rules are in the hashing and standardization guide.
ZIP codes
1. Leading zeros kept on the ZIP
The technical specification strips leading zeros from the ZIP before it keeps the first five characters. So 02134 is hashed as 2134. Code that keeps the zero, which is what a plain reading of “the first 5 characters” in the regulations gives, misses every ZIP that starts with 0. In the US those are ZIPs in New England, New Jersey and Puerto Rico.
input "02134"
wrong "02134" → H3Xi2sW4JSHx…
right "2134" → 2qs6poGFtne9…The ZIP only feeds the NDZ list, so the visible effect is on the composite. For Jane Doe, born 22 July 1990, the NDZ built with the kept zero is 85i1uzDPoLZy…; DROP's is o9xU5XHeI9i5….
Strip leading zeros after removing punctuation and before taking five characters. A value of all zeros becomes 0. The regulations and the November 2026 proposed rules describe the order differently from the specification, and that difference is not settled. DROP builds its hashes with the specification, so we follow it and will update this page if the final rule text changes it.
2. ZIP+4 suffix left on, or a Unicode dash not recognized
The specification drops the separator and the last four digits when the whole value is a ZIP+4. For most ZIPs the five-character cut hides a missed step. It shows when the ZIP starts with 0: once the zero is stripped, a digit of the +4 moves into the five characters. The same happens when the separator is an en dash or another Unicode hyphen that an ASCII-only pattern does not see. Version 1.3.0 counts U+2010 to U+2014, the minus sign U+2212 and the full-width plus as separators.
input "02134–1234"
wrong "21341" → vhqkkEkvckxM…
right "2134" → 2qs6poGFtne9…Recognize every listed separator. If your ZIP column stores nine digits with no separator, the specification has no rule to drop the +4, so hash both the nine-digit value and its first five digits; that is the kind of extra step § 7613(a)(1)(A) asks for.
Names
3. Names not NFKC-normalized, or format characters left in
Names start with Unicode NFKC normalization, which turns full-width letters into ordinary ones. Then control and format characters are removed. Zero-width joiners and spaces pasted from web forms are format characters, and so is a byte-order mark stuck to the first value of a CSV file. Full-width letters are still letters, so a letters-and-digits filter keeps them.
input "Jane"
wrong "jane" → qSnoy8Act8Tu…
right "jane" → gfj23eiDZfOS…Run NFKC first, then remove Unicode categories Cc and Cf, then lowercase.
4. Transliteration table skipped, or run after accent folding
DROP publishes a transliteration table for Greek, Cyrillic and special Latin letters: ø becomes o, ł becomes l, ß becomes ss. These letters do not decompose under Unicode normalization, so generic accent folding leaves them in place. An ASCII-only filter goes the other way and deletes them.
input "Søren"
wrong "søren" → a+oBRwFa/L54…
wrong "sren" → +58aNuj5N32z…
right "soren" → 99LKojRvJCNr…Apply DROP's table after lowercasing and before folding accents (María → maria), and keep letters from scripts the table does not cover. Chinese, Hebrew and Arabic names are hashed in their own script.
5. Spaces, hyphens and apostrophes left in names
After transliteration and folding, a name keeps only letters and decimal digits. Apostrophes, hyphens, periods and spaces all go, and compound names stay in their own field as one word.
input "O'Connor"
wrong "o'connor" → Ty70pqQ/Jpeb…
right "oconnor" → eg+/30DL65dC…input "Juan Pablo"
wrong "juan pablo" → tre7darvk0p0…
right "juanpablo" → 91hIbrbzNeqH…Filter to Unicode letters and decimal digits as the last name step. Do not move a middle name into the last name field.
Dates of birth
6. Date of birth not in YYYYMMDD
DROP hashes the birth date as eight digits: four-digit year, two-digit month, two-digit day, no separators. An ISO date with hyphens, a US date with slashes or a database timestamp each hash differently. Consumers can give DROP only one birth date, so there is no second chance on another row.
input "1990-07-22"
wrong "1990-07-22" → JmSdXQudxBUJ…
right "19900722" → 0oJHwSqomfyh…Inside the NDZ the wrong date gives mwcC6YgOwpAE… where DROP has X7qUgPCyldid…. A quieter version of the same slip is day and month swapped: 04/07/1985 is 19850407 in US order and 19850704 in day-first order, and only one of them is the person's birthday.
Convert to YYYYMMDD from a known source format. Refuse two-digit years rather than guess the century.
Phone numbers
7. Phone extension or country code left in
The phone rule keeps digits only and then the last 10. Two things break it. An extension stored in the same field adds digits at the end, so the last 10 shift; DROP says it does not remove extensions for you. And a number written in E.164 form keeps its country code if you hash all the digits.
input "+1(212)555-0186 ext. 123"
wrong "5550186123" → f3TE4Q75X5lM…
right "2125550186" → IEWV38qOhXiz…input "+14155559317"
wrong "14155559317" → XIPRAOBijcVn…
right "4155559317" → vGM7y5n+hBXR…Remove the extension marker and its digits first, then keep digits, then take the last 10. Do not add back a country code stored in another column. DROP's only extension example is ext. 123, so check your own data for other markers such as x or #.
Email addresses
8. Email cleaned too little, or too much
The email rule is short: remove all whitespace, including inside the address, and lowercase. Everything else stays, including dots and plus tags. A missed lowercase step is enough to break the match. The opposite slip is Gmail-style folding of dots and plus tags.
input "[email protected]"
wrong "[email protected]" → PpmRlTqsc/ik…
right "[email protected]" → KA18MT/ph6IH…input "[email protected]"
wrong "[email protected]" → 1hFzBkhe0OUK…
right "[email protected]" → 9BQ5yUSnY8mB…Strip whitespace and lowercase, nothing more. If you hold both the folded and the original address, hash the original.
The hash step itself
9. Hex or URL-safe Base64 instead of standard Base64
Most SHA-256 functions return hex by default. DROP uses standard Base64 with padding: 44 characters, the alphabet with + and /, ending in =. The digest bytes are the same; only the text differs, so no string comparison will ever succeed. Here is the same digest of [email protected] written three ways:
hex (wrong) 280d7c313fe987a207623cd3f73c044f2483432bd287ced8ba8481a4e42d9211
URL-safe (wrong) KA18MT_ph6IHYjzT9zwETySDQyvSh87YuoSBpOQtkhE
standard (right) KA18MT/ph6IHYjzT9zwETySDQyvSh87YuoSBpOQtkhE=Encode the 32 digest bytes as standard Base64 with padding. A quick check: every DROP hash matches ^[A-Za-z0-9+/]{43}=$.
10. Composite hash built from the wrong pieces
NDZ and NameVIN are hashes of hashes. Hash each standardized field, join the Base64 strings in order (first name, last name, then date of birth and ZIP, or VIN) with their trailing = and no separator, and hash the joined string. Each of the three shortcuts below gives a different result. This is CalPrivacy's webinar example, Jane Doe, 7/22/1990, 94103-3857:
raw "janedoe1990072294103" → Cr1bAcEfrgd/…
bytes "digest bytes joined" → 0iLTT39TDGGf…
comma "hashes joined with ," → BrzPBd1vLZA3…
right "hashes joined as text" → X7qUgPCyldid…Build it as base64(sha256(h(first) + h(last) + h(dob) + h(zip))), where each h is the 44-character Base64 string. Every part must be present; there is no NDZ without a ZIP.
11. Bytes that are not UTF-8
DROP hashes the UTF-8 bytes of the standardized value. Some environments hand the hash function other bytes. SQL Server's HASHBYTES on an NVARCHAR column hashes UTF-16, which changes every value, ASCII included. Older Latin-1 or Windows-1252 exports change only values with non-ASCII characters, such as names kept in their own script.
input "张"
UTF-16 "张" → C1v5EQse3lBJ…
UTF-8 "张" → 9WoEn8wj8ev8…input "[email protected]"
UTF-16 "[email protected]" → oWYirpJm9Yxc…
UTF-8 "[email protected]" → KA18MT/ph6IH…Encode to UTF-8 explicitly before hashing, or cast to a UTF-8 or single-byte type for ASCII-only data, and test with a non-ASCII value.
Vehicle and device IDs
12. VIN, MAID or CTVID left in its stored case or format
VINs are usually stored in capitals and advertising IDs with hyphens. DROP lowercases all three identifiers, removes spaces and hyphens from VINs, keeps only hex characters in a MAID (which must then be exactly 32), and keeps letters and digits in a CTVID (8 to 32 of them).
input "1HGCM82633A004352"
wrong "1HGCM82633A004352" → hLEZVtpzsvBa…
right "1hgcm82633a004352" → iNswy1m+0VSt…input "A3F1C2D4-5678-90AB-CDEF-1234567890AB"
wrong "a3f1c2d4-5678-90ab-cdef-1234567890ab" → Stwc1bCPnkQS…
right "a3f1c2d4567890abcdef1234567890ab" → 250KY6lOgzYU…Apply each list's own rule, and treat a value that fails it as having no hash. A VIN with a slash, such as 1HGCM/82633A004352, is invalid under version 1.3.0, and so is a 30-character MAID. Hashing whatever is left over would never match.
All 12 in one table
| # | Slip | List affected | Symptom | Fix |
|---|---|---|---|---|
| 1 | Leading zeros kept on the ZIP | NDZ | No NDZ match for ZIPs starting with 0 | Strip zeros, then keep 5 characters |
| 2 | ZIP+4 left on, or Unicode dash missed | NDZ | Misses on zero-led ZIP+4 values | Drop the +4 for every listed separator |
| 3 | No NFKC, format characters left | NDZ, NameVIN | Misses on full-width or pasted names | NFKC, then remove Cc and Cf |
| 4 | Transliteration skipped or late | NDZ, NameVIN | Misses on ø, ł, ß, Greek and Cyrillic names | DROP's table before accent folding |
| 5 | Punctuation or spaces in names | NDZ, NameVIN | Misses on O'Connor, Juan Pablo, Mary-Ann | Keep letters and digits only |
| 6 | Birth date not YYYYMMDD | NDZ | Few or no NDZ matches | Eight digits from a known format |
| 7 | Phone extension or country code kept | Phone | Misses on E.164 numbers and extensions | Remove extension, digits, last 10 |
| 8 | Email cleaned too little or too much | Misses on capitals, or on dotted Gmail | Remove whitespace, lowercase, nothing else | |
| 9 | Hex or URL-safe output | All | Zero matches on every list | Standard Base64 with padding |
| 10 | Composite from the wrong pieces | NDZ, NameVIN | Zero composite matches | Hash the joined Base64 field hashes |
| 11 | Bytes not UTF-8 | All, or non-ASCII only | Zero matches, or misses on non-Latin names | Encode as UTF-8 before hashing |
| 12 | VIN, MAID or CTVID format kept | NameVIN, MAID, CTVID | Zero matches on that list | Lowercase and apply each list's rule |
How do I test my DROP hashing before a real batch?
Run CalPrivacy's published examples through your routine and compare every character. Our vector check does this on every build of this site: today 15 of 15 examples reproduce, including two the documentation prints incorrectly. They are listed with the correct hashes in the test vector table.
The published examples do not cover every slip above. Add your own cases for a ZIP with a leading zero, a ZIP+4 with an en dash, a full-width name, a name with ø or ß, and a phone number with an extension. The free DROP hash checker shows the standardized value and the hash for any input, computed in your browser; nothing you type leaves the page.
If a whole list comes back with almost no matches, suspect the hash step (slips 9 to 11) before the data. If one list matches well and NDZ does not, look at ZIP and date handling first.
Questions
Why don't my DROP hashes match my customer records?
Look first at standardization. One step that differs from CalPrivacy's rules is enough: a ZIP keeps its leading zero, a name keeps punctuation, an email keeps capitals, or the hash is written in hex instead of standard Base64. One changed character gives a completely different hash, and the work item is reported as status 5 (not found).
Does DROP strip leading zeros from ZIP codes?
Yes, under the technical specification (version 1.3.0). Leading zeros are removed before the first five characters are kept, so 02134 is hashed as 2134. The regulations word the step differently, and that difference is not settled; DROP builds its hashes with the specification.
Should I remove dots and plus tags from Gmail addresses before hashing?
No. DROP's email rule removes whitespace and lowercases, and keeps dots, plus signs and every other character. [email protected] and [email protected] are different hashes.
How can I check my hashing is right?
Reproduce CalPrivacy's published test vectors exactly. Our vector check reproduces 15 of 15 on every build; the free hash checker at /tools/drop-hash-checker shows the standardized value and hash for any input you type.
More from the blog
- What 625 data brokers told California: the 2026 registry in numbers
- DROP deletion requests vs CCPA deletion requests
- The 2027 DROP calendar for the 30-day cycle
Sources
- CalPrivacy: DROP operations and API documentation (hashing rules, transliteration, test vectors)
- CalPrivacy: DROP technical reference (status codes, version history)
- CalPrivacy: Processing DROP requests
- DROP regulations, 11 CCR §§ 7600–7622 (PDF)
- CalPrivacy: DROP audits (proposed regulations, November 2026)
Written by the CalDrop Compliance team at Preptics LLC and last reviewed 22 November 2026. This guide explains the rules; it is not legal advice. CalDrop Compliance is an independent product of Preptics LLC. It is not affiliated with, endorsed by or operated by the State of California or the California Privacy Protection Agency (CalPrivacy).
