Feature Data Privacy
The 45-Day DROP Clock Is a Batch Job Now. Build the File That Proves It Ran.
California's DROP turned deletion into a recurring 45-day batch with $200/day per-request penalties. Here's the evidence file that survives review.
Table of Contents
CalPrivacy has spent two years building a paper trail against data brokers for not registering, and it has started issuing orders that hit revenue, not just fees: Rickenbacher Data (d/b/a Datamasters) was fined $45,000 in January 2026 and ordered to stop selling Californians’ personal information entirely; S&P Global paid $62,600 the same day for what the agency described as an administrative error. Neither was a horror story about data quality. Both were paperwork failures that a competent privacy ops function would have caught.
DROP (the Delete Request and Opt-out Platform) is where that fight moves next. As of August 1, 2026, brokers must access the platform at least once every 45 days and process every request in the batch. The penalty for a failure to delete is $200 per day, per request, plus the agency’s investigation and administrative costs.
That arithmetic is the whole story. A 12-day slip on a batch of 4,000 requests is $9.6 million of theoretical exposure before anyone argues about scope. In practice you won’t eat all of it — but the shape of the risk is that your unit of failure is a single consumer request, multiplied by days, and the evidence that resolves it is a machine log.
Two clocks, and people keep merging them
The first compliance design flaw is treating this as one deadline. It’s two.
| Clock | What it measures | What starts it | Where it lives |
|---|---|---|---|
| Access cadence | Days between pulls of the DROP list | Date of the last download | DROP account audit trail |
| Per-request disposition | Days between a request entering your copy of the list and its terminal status | The request’s own timestamp | Your ticketing system + DROP status reporting |
Missing the cadence clock means requests are aging somewhere you can’t see. Missing the disposition clock means you pulled on time and still blew the deadline. Both are $200 per day per request, and neither is cured by a well-written policy.
The trap is that the cadence is easy to prove and the disposition is not. A team under pressure will happily produce a screenshot of an on-time download and call the control demonstrated. An examiner who knows the statute will ask what happened to request #3,881.
The evidence file, built before you need it
Stop thinking of this as deletion work. Deletion is the byproduct. The deliverable is a defensible run record. Six artifacts, all of which should exist whether or not anyone ever asks.
| Artifact | What it has to contain | What a bad version looks like |
|---|---|---|
| Run log | Run ID, cadence date, download timestamp, file hash, row count, operator | A calendar invite and a shared-drive folder named “DROP” |
| Hash/standardization spec | Field-by-field normalization rules (case, punctuation, +1 stripping, apartment-line handling), hash algorithm, salt custody | ”We hash it” with no spec, so two runs process differently |
| Match report per run | Rows received, rows matched, rows unmatched by reason, match rate | A total match count with no denominator |
| Disposition register | One row per request: matched record ID, action, terminal status, status-reported date | A single “batch completed” note |
| Exception log | Unmatched reason, second-pass method, outcome, owner, close date | Unmatched rows silently dropped at export |
| Cadence attestation | Who confirms the next pull date, and the backup who covers PTO | One named person, no backup |
The salt custody item is the one teams skip and then regret. If the hash salt rotates without version tagging, last quarter’s match report stops being reproducible, and “we matched 98%” becomes an assertion rather than a measurement.
A realistic hypothetical: a 2.1 million-row list, 61% match rate on first pass, 71% after the second pass on phone plus ZIP. The remaining 29% is where the penalties are — and the reason the exception log, not the match rate, is the artifact that decides whether the run holds up.
Matching is the hard part, and it’s where the money is
Broker data arrives dirty by design. Lists sold as “verified” carry nicknames, unit numbers glued to street lines, shared family emails, and phone numbers recycled after a decade. Your identity resolution has to handle a consumer who opted out using “Bob” when your record says “Robert.”
Three practical rules that keep matching defensible:
- Two deterministic passes minimum, with method documented per pass. Pass one on hashed email plus hashed phone. Pass two on hashed name plus ZIP plus a hashed secondary identifier. Log which pass produced each match.
- Never let fuzzy matching close a record without a confidence tier. “Probable” matches need an owner decision recorded, and probable-equals-deleted is a defensible default that beats probable-equals-dropped.
- Cap the list size per run so a spike doesn’t blow the clock. If the 30-day cadence arrives and your monthly transfer volume has doubled, a run sized to last month’s volume will age out mid-batch.
Verification failures are not a parking lot
DROP’s own guidance walks brokers through download, standardize, hash, match, process, report. The report step is a status write-back inside 45 days. What the workflow diagram undersells is the request that matches three records with different addresses, or matches nothing at all.
The Delete Act gives you the answer: where a broker denies a request because it cannot verify it, the request is processed as an opt-out of sale instead. That is a disposition, not a dead end. Record the downgrade with the reason code and move on. A request sitting in a “needs review” queue on day 46 is already accruing.
CalPrivacy also signaled where it looks. Its Enforcement Division launched a Data Broker Enforcement Strike Force in November 2025 specifically to review Delete Act compliance alongside CCPA compliance, and the January 2026 orders show the agency will accept an audit-procedure remedy — S&P Global had to adopt registration and compliance-auditing procedures. That’s the tell: they want the control, and they will write one for you if you don’t have it.
Thirty days arrives January 1, so build for 30 now
AB 883, signed September 27, 2026, replaces 45 days with 30 effective January 1, 2027, and adds separate relief for judges and elected officials whose information a broker holds. Any process tuned to clear 45 by a day or two is functionally late under 30.
Rebuild on this sequence, and put dates on it:
- Week 1: Move the run calendar to a 26-day internal cadence so 30 is a backstop rather than a target.
- Week 2: Re-derive list-size ceiling from the highest single transfer in the last 90 days, plus 40%. Set an alert at 80% of the ceiling.
- Week 3: Automate the unmatched-row downgrade so verification failures close as opt-outs instead of waiting for a human.
- Week 4: Run one full parallel cycle at 30 days with the new exception automation and time the close-out of every row.
- Week 6: Tabletop the scenario that breaks this — a payment-processor outage on day 28, or a hash salt rotation mid-batch.
Cutting the latency in half is mostly a matter of removing human handoffs from steps that don’t need judgment: file intake, normalization, first-pass matching, and status write-back. Keep people on ambiguous matches and on the exception review that decides whether the automation is degrading.
So What?
California has made data deletion a recurring, machine-checkable batch job with a per-request daily meter. That is a different control type than a DSAR process, and if you’re still staffing it like one, the gap will show up as a cadence miss rather than a wrong answer. Look at three numbers this week: days since your last pull, count of open rows older than 30 days, and whether your run log would survive a request for the download timestamp and file hash. If the third one is soft, the first two don’t matter.
The Compliance Essentials bundle has the policy, control, and evidence skeletons that sit around this workflow.
FAQ
Does the 45-day clock run from the batch or the request?
Treat it as per-request. The statute is written around processing requests after receipt, and the agency’s guidance frames it as access at least once every 45 days plus a disposition reported within 45 days. If you have to choose, the stricter reading is the one that survives an exam.
Are we a data broker if we sell data about our own customers?
Usually not. The definition requires knowing collection and sale of personal information about consumers with whom the business has no direct relationship. Customer data you sell to a partner sits outside the core definition — but the “no direct relationship” test is applied to the consumer whose data it is, not to your commercial counterparties, so acquired lists and co-op datasets need a separate look.
What’s the cheapest way to prove we didn’t miss a run?
A timestamped DROP account trail plus a file hash per download. Both are generated as a side effect of doing the work, cost nothing to keep, and are the two facts that resolve most cadence disputes. The expensive part is reconstructing them after the fact.
Related reading
◆ Need the working template?
Start with the source guide.
These answer-first guides summarize the required fields, evidence, and implementation steps behind the templates practitioners search for.
◆ Related template
Compliance Essentials
Multi-domain compliance coverage: data privacy, incident response, BCP/DR, and SOC 2 — 42% off.
◆ Immaterial Findings · Weekly
Sharp risk & compliance insights. No fluff.
◆ FAQ
Frequently asked questions.
What is DROP and who has to use it?
Does the 45-day deadline apply to each request or to the whole batch?
What changes on January 1, 2027?
What happens if a request can't be matched to a record?
Author
Rebecca Leung
Rebecca Leung has 8+ years of risk and compliance experience across first and second line roles at commercial banks, asset managers, and fintechs. Former management consultant advising financial institutions on risk strategy. Founder of RiskTemplates.
◆ Related framework
Compliance Essentials
Multi-domain compliance coverage: data privacy, incident response, BCP/DR, and SOC 2 — 42% off.
◆ Keep reading
Related posts.
Data Privacy
GM Paid $220 Million for Consent Its Dealers Captured. Your Consent Management Has the Same Hole.
The 36-state GM settlement turns delegated consent capture and vendor data assessments into enforcement items. Consent management controls for dealer, agent, and partner channels.
Nov 20, 2026
Data Privacy
Consent Receipts for Sensitive-Data Sales Must Survive Five Years. Most Records Are Two Fields.
Delaware HB 380 requires a five-year consent record produced with the DPIA. Here is the field set, the versioning rule, and the deletion conflict.
Nov 16, 2026
Data Privacy
Neural Data Becomes Sensitive Data in Delaware on January 1. Run the Signal Test First.
Delaware HB 380 makes neural data sensitive data on Jan 1, 2027, joining California, Colorado, Connecticut and Montana. Here is the measured-vs-inferred test and the vendor evidence.
Nov 16, 2026