The Quarterly Dead-Account Sweep: Company Book to Six Proven Outcomes
Run a dated sweep that gives every CRM account one evidence-backed outcome, applies the matching write-back, and ends with counts a second operator can match.
Key takeaways
- A lapsed LEI is not evidence of closure: the issuer must confirm the entity is still operating before it sets LAPSED, so the status leans toward Live.
- A domain that enters its 30-day redemptionPeriod is about 35 days from being registrable by anyone, so a quarterly sweep can miss a full drop and re-registration unless it compares the domain's RDAP creation date to the account's created date.
- The Companies House free bulk product lists only live companies and the dissolved-search API covers dissolutions only from 2010 onward, so a company missing from the bulk file is not proven dissolved.
- In HubSpot, when two company records share a domain, new contacts with that domain go to only one of them and you cannot choose which, so remove the domain from the dead record before you give it to the successor.
- Salesforce keeps merged-away accounts in the Recycle Bin for 15 days, documents no unmerge, and loses unique fields you did not choose to keep, so export before any merge and link by default.
- In Refolk's index, US RevOps profiles tagged HubSpot outnumber Salesforce-tagged ones 183 to 66 (2.8x), so HubSpot's shared-domain routing rule is the misrouting mechanism most teams will hit.
This playbook is for the RevOps or recruiting-ops owner who has to find, every quarter, the company records in the CRM that point at businesses that closed, merged away, or lost their domain. It starts from a full account export and gives every account one evidence-backed outcome. It then applies the matching write-back without losing deal history or misrouting new leads, and ends with a before-and-after count that a second operator can re-run and match.
What are the six outcomes of a dead-account sweep?
Every account leaves the sweep with exactly one outcome: Live, Renamed, Absorbed, Dissolved with a trading successor, Defunct, or Domain Lost. The outcome is the decision. The evidence behind it and the write-back after it both follow from which one you pick.
Two terms first. A legal entity is the registered company that a registry tracks by number. A trading business is the operation that sells, hires, and answers email. Usually they are the same thing. This sweep exists because sometimes they are not.
| Outcome | What it means | Write-back |
|---|---|---|
| Live | Entity active, domain still the company's | Stamp sweep date only |
| Renamed | Same registry number, new name or domain | New domain primary, old one aliased |
| Absorbed | Entity ended into a named successor | Link or merge to successor record |
| Dissolved, trading successor | Shell entity gone, business trades on under another entity | Update legal fields, keep account |
| Defunct | Entity ceased, no successor, nothing trades | Archive with history, suppress contacts |
| Domain Lost | Domain left the company's control, new one unconfirmed | Strip domain, hold for review |
Precedence matters because "one outcome" is the rule. If the entity evidence proves Absorbed or Defunct and the domain has also dropped, assign the entity outcome and strip the dropped domain as part of that outcome's write-back. Domain Lost is for rows where the domain is gone and the entity evidence does not yet say what happened. If you later confirm the company's new domain under the same registry number, the row becomes Renamed.
The hardest call is telling a dissolved shell from a business that still trades. The matrix below is how I sort them.
Shell entity or trading business
The top-right quadrant splits in two. Absorbed means a corporate action ended the entity and a successor is named, either in the registry or in the LEI SuccessorEntity field. The account's history now belongs with a larger record. Dissolved with a trading successor means the entity on your record dissolved but the same business carries on under a different entity, with no recorded merger. The customer relationship has not changed. Only its legal identity has, so you update the legal fields and keep the account.
Which evidence comes first, and what does each source prove?
Read the registry status by company number first. Then read the LEI registration status and successor fields, then domain status and redirects, and finally web and archive captures as corroboration. Registries settle the legal outcome, domains show the routing risk, and archives only date what the earlier layers already suggest.
No public source sets this order. Vendor content recommends a 90-day CRM hygiene cadence but does not say which evidence to read first, and whether registries or domains should come first is still unsettled in public sources. I put registries first because they are the only source whose job is to record legal status. The domain check still runs on every row whatever the registry says, because a live company can lose its domain.
Order of evidence in the sweep
- Registry status by numberIs the legal entity active, dissolved, cancelled, or renamed
- LEI recordDid the entity retire, and is a successor entity named
- Domain status and redirectDoes the company still control its domain, and where does it point
- Web and archive capturesWhen did the site change, as a cited date for a non-Live outcome
Registry status: what it proves and where it lies
Companies House (UK). The dissolved-search endpoint is GET https://api.company-information.service.gov.uk/dissolved-search/companies, with alphabetical, best-match, and previous-name search modes. It returns at most 100 results per page. Coverage of dissolved companies starts on 1 January 2010, and statutory records are archived 20 years after dissolution. The free bulk Company Data Product holds basic data on live companies only. A Companies House staff reply confirms there is no dissolved bulk product. So to check dissolved status in bulk for UK accounts, you loop the API by company number. You cannot diff against the bulk file.
- Proves: the entity's legal status, and a name change through previous-name search.
- Lies when: a company is missing from the live snapshot because of a name or number mismatch, not because it dissolved.
Ohio. The cancelled/dissolved corporations list covers the last five years and is updated weekly. Status values are Active, Hold, Canceled, and Dead, and the list also shows PermanentlyCancelled. The broader monthly business reports are generated on the second Saturday of each month. Entities can be reinstated up to two years after the cancellation date, and a reinstated entity stays on the list until the next weekly update.
- Proves: a cancellation or dissolution within the lookback window.
- Lies when: the entity has been reinstated since the last refresh, or it closed more than five years ago and so is absent from the list.
Other states and countries. Each registry publishes on its own terms. Before you trust a list, record its lookback, refresh cadence, and status vocabulary in the evidence log, and treat any of the three you cannot confirm as unknown.
LEI records: what they prove and where they lie
An LEI (Legal Entity Identifier) is a global identifier for legal entities, and its record carries a registration status and, from format 3.1, a repeating SuccessorEntity field. The table maps each status to what it means for the sweep.
| RegistrationStatus | Official meaning | SuccessorEntity | Sweep implication |
|---|---|---|---|
| ISSUED | Current | Omitted | Weak evidence of Live |
| LAPSED | Not renewed, not known to have ceased | Omitted | No closure evidence |
| RETIRED | Ceased, not merged | May be present | Defunct or Absorbed, by successor |
| MERGED | Deprecated in LEI-CDF 3.1 | Required in CDF 2 | Legacy records only |
The GLEIF data quality checks say SuccessorEntity can be present when the status is RETIRED or ANNULLED, and is omitted when it is ISSUED, LAPSED, PENDING_TRANSFER, or PENDING_ARCHIVAL. EntityExpirationDate and EntityExpirationReason are deprecated too. Their old reason values were DISSOLVED, CORPORATE_ACTION, and OTHER, so older records may still carry them. One public scraper reports that no record in the index currently carries MERGED. Treat that as a secondary observation, not a guarantee. I have not verified how often GLEIF refreshes, so record the retrieval date of every LEI you read.
When SuccessorEntity is filled, EntityStatus shows the last status before the record was superseded, which is not necessarily its status today. An Absorbed outcome therefore needs the successor's own record fetched and read separately.
A lapsed LEI is weak evidence the company is alive, not evidence that it died.
Domain and website: what they prove and where they lie
RDAP is the registration data lookup that returns a domain's status codes and dates. EPP status codes are the standard flags registries set on a domain. For each domain the sweep needs three facts: the status codes, the registration creation date, and the final host after following HTTP redirects.
- redemptionPeriod: a 30-day window for gTLDs, required by ICANN's Expired Registration Recovery Policy. The domain has expired, but the original owner can still recover it.
- pendingDelete: about five days and cannot be reversed. Once the domain is deleted, anyone can register it.
- serverHold: usually applied because of legal disputes, court orders, or UDRP proceedings. It does not mean the company closed.
- Creation date later than the account's created date: the domain was registered again after your record was made. That is the clearest sign a new owner holds it.
- Cross-domain redirect: the site now points somewhere else. On its own this cannot tell a rebrand from an acquisition.
Web and archive captures
How long it takes certificate-transparency lapses, parking pages, and gaps in web-archive captures to show up has not been established publicly. Use captures for one job: citing a dated observation for every non-Live outcome. A capture never settles an outcome by itself.
How long does each signal stay valid?
Every signal has a shelf life, and a sweep that reads a stale signal assigns the wrong outcome. Record the snapshot date next to every status, and re-check any signal whose window has closed by the time you write back.
| Signal | Window | What it means for the sweep |
|---|---|---|
| Domain redemptionPeriod, then pendingDelete | 30 days, then about 5 | Can drop and be re-registered within one quarter |
| Ohio cancelled/dissolved list refresh | Weekly | Up to a week stale |
| Ohio list lookback | 5 years | Older closures need the entity record |
| Ohio reinstatement window | 2 years | Dead or Canceled can come back |
| Companies House dissolved coverage | Since 2010; archived after 20 years | Older dissolutions are not in the API |
| Salesforce merge recovery | 15 days in Recycle Bin | No documented unmerge after that |
The domain row matters most because it is shorter than the sweep cycle. The flow below shows how a quarterly run can miss a whole drop.
How a dropped domain reaches a dead account
- ExpiryThe registration is not renewed and the domain enters redemptionPeriod
- Redemption30 days in which the original owner can still recover it
- pendingDeleteAbout 5 days, and it cannot be reversed
- DropThe domain is deleted and anyone can register it
- Re-registrationForm fills and auto-created contacts from the new owner attach to the dead account
How do you run the dead-account sweep, step by step?
Run nine steps in a fixed order, starting from a frozen export and ending with a second operator's matching count. The durations below are my desk estimates, not published benchmarks. Expect them to grow with account volume and the number of registries you touch.
The quarterly dead-account sweep
- Freeze a dated exportExport every account with record ID, primary and secondary domains, legal name, registry number, LEI, and the channel that created the record. Log the date and row count.
- Match each account to its registryLook up each row by registry number in the Companies House dissolved search, the Ohio list, and the relevant state lists. Record the status and the source's snapshot date.
- Resolve LEIs and fetch successor recordsRead RegistrationStatus and SuccessorEntity, then fetch each successor's own record. Record the successor LEI and name, or an explicit null.
- Check domain status and redirectsScript RDAP status codes, creation date, and the final host after redirects for every primary and secondary domain.
- Corroborate with web and archive capturesCite a dated capture for every row heading to a non-Live outcome. Never use a capture as the only evidence.
- Assign one outcome per accountThe analyst proposes, a reviewer confirms, and each outcome has at least one dated evidence row. Ambiguous rows go to the exceptions log.
- Apply the write-back for each outcomeAlias or strip domains first, then link or merge, update legal fields, suppress, or archive. Export any record before merging it.
- Rectify or suppress contactsWith the privacy owner, re-point contacts who verifiably moved and suppress the rest, keeping a minimised suppression record.
- Reconcile with a second operatorA second operator re-runs the sweep from the frozen export and evidence log. Done when the before-and-after counts per outcome match.
Owners and rough time per step:
- Export: RevOps owner, about 1 hour. Freeze it first because nothing says related records move back after a merge. This file is your only clean "before".
- Registries: ops analyst, 2 to 6 hours. Match by number. Name matching is how false dissolutions get in.
- LEIs: analyst, 1 to 2 hours. Fetching the successor record is the step people skip.
- Domains: a script, under 1 hour. Include secondary domains. A forgotten alias can be the one that dropped.
- Captures: analyst, time varies. Cap the effort per row and send the overflow to the exceptions log.
- Outcomes: analyst, then a reviewer who did not do the matching.
- Write-back: CRM admin, following the order in the next section.
- Contacts: ops plus the privacy owner.
- Reconcile: a second operator who did not run the first pass.
What write-back does each outcome get in HubSpot and Salesforce?
Order matters more than the individual edits. Remove or demote the dead record's domain before you link or merge, and export before any merge, because neither CRM gives you a clean undo. The rest of this section is the mechanics behind that rule.
HubSpot
Four documented behaviors drive most HubSpot misrouting:
- Deduplication uses only the primary Company domain name, not secondary domains.
- Companies created through the API are not deduplicated by domain.
- When two companies share a domain, new contacts with that domain go to only one of them, and you cannot choose which.
- The duplicates manager needs a Professional or Enterprise subscription.
What follows for each outcome:
- Renamed: make the new domain primary and keep the old one as a secondary alias on the same record.
- Absorbed: remove the domain from the dead record before adding it to the successor. If both records hold it, HubSpot decides which record gets new contacts, not you.
- Dissolved, trading successor: leave the domain where it is if the trading business still uses it, and update the legal name and registry number.
- Defunct: remove the domain, then archive. A Defunct record that keeps a live-looking domain keeps collecting contacts.
- Domain Lost: remove the domain from the record entirely. It now belongs to someone else.
Salesforce
A Salesforce merge re-parents Opportunities, Contacts, Activity, Campaign History, and Cases to the surviving account. It handles up to three accounts at a time. Merged-away accounts sit in the Recycle Bin for 15 days, and Salesforce Help documents no unmerge. Unique data on the deleted records that you did not choose to keep is lost.
My default for Absorbed is a link: a successor lookup field on the dead account pointing at the successor account, with the dead account archived. That keeps the old record's history auditable and lets you undo a wrong Absorbed call. Merge only when the dead record holds nothing unique and you have already exported it. For Defunct, archive with history kept, so the deal history stays attached to a record a reviewer can find.
Re-pointing people after an Absorbed outcome
Linking the account is the easy half. The harder question is which of the old team now works at the successor, because re-pointing a contact you have not verified just moves an inaccurate record to a new account. Refolk answers that as a plain-English search across LinkedIn, GitHub, and the open web, so the re-point rests on a current employer rather than an assumption.
Which CRM's rules to learn first
| Query in Refolk's index | Count | Example employers |
|---|---|---|
| US RevOps, Salesforce tag | 66 | Autodesk, Qualys, Nextdoor, Metabase |
| UK RevOps, Salesforce tag | 13 | Kpler, BigHand, The Citation Group, VAST Data |
| US RevOps, HubSpot tag | 183 | Panorama Education, doxy.me, Merchants Fleet |
These counts depend on explicit skill tags, so they are lower bounds, and the HubSpot sample also picked up adjacent titles such as Renewals and GTM Strategy. Two things follow. First, HubSpot's shared-domain routing is the misrouting mechanism most teams will meet, so learn its rules before Salesforce's. Second, the UK Salesforce pool is 13 against 66 in the US, a 5.1x gap. If you need a second operator with UK Salesforce experience for step 9, line them up before the quarter starts.
How does the dead-account sweep go wrong?
Most wrong outcomes come from reading a signal as proof of something it does not prove: a lapse taken as closure, a missing row taken as dissolution, a redirect taken as an acquisition. Each failure below gives what it looks like, why it misleads, and the check that catches it.
- LAPSED read as dead. The LEI shows LAPSED and the account gets marked Defunct. A lapsed LEI does not mean the company was dissolved, and the issuer confirmed the entity was operating before it set the status. Check: read the registry status. Treat LAPSED as neutral.
- Missing from the Companies House live snapshot read as dissolved. A company is absent from the bulk file, so someone assumes it dissolved. Name or number mismatches produce exactly this false positive, and the bulk file has no dissolved counterpart. Check: query the dissolved-search API by company number.
- Ohio Dead or Canceled treated as final. The weekly list shows Dead and the account is archived. Entities can be reinstated for up to two years, and the list lags reinstatements by up to a week. Check: re-read the entity record before write-back. Put a re-check date on every Ohio Defunct inside the reinstatement window.
- serverHold read as closure. The domain is on hold and the account is marked Defunct. serverHold usually follows a legal dispute, a court order, or a UDRP proceeding. Check: confirm whether the trading entity is still filing with its registry.
- Rebrand read as acquisition. The old domain redirects to a different domain and the account is marked Absorbed. A cross-domain redirect alone cannot tell the two apart. Check: run the registry previous-name search. If the registry number is unchanged, it is Renamed.
- Dropped domain left as primary. The account looks Live in the CRM because new contacts keep arriving. After about 35 days a new owner can register the domain, and their form fills and auto-created contacts attach to your dead account. Check: compare the RDAP creation date with the account's created date. A later creation date means a new registration.
- HubSpot API integrations recreate duplicates. The cleaned domain is back on a fresh record next quarter. API-created companies bypass domain dedupe. Check: filter companies created after the sweep by record source, and fix the integration, not just the record.
- Salesforce merge wipes unique fields. The surviving account is missing data the dead one held. Unique data on deleted records that was not chosen for retention is lost, and the 15-day Recycle Bin does not bring back merged-in field values. Check: export every record before you merge it.
Two benchmark traps sit next to these. The widely repeated 22.5% annual decay figure traces back to a study nobody has found, and it is about a decade old. Vendor claims of 30 to 40% a year at startups and VC-backed companies come without published methods. Closure statistics undercount account churn the other way. A secondary source citing US Census data puts 2023 closures at about 700,000 locations, 9.4% of employer establishments, but Census does not count M&A as firm death. That makes 9.4% a floor for Absorbed plus Defunct, not an estimate of it.
What happens to contacts at a dead account?
Re-point the contacts who verifiably moved, suppress the rest, and keep only a minimised suppression record. Do not keep the full contact record once its employer is gone. The account outcome tells you which contacts to look at. It does not settle what you may keep.
The relevant UK duties:
- Accuracy: inaccurate data must be erased or rectified without delay, and the ICO checklist expects controllers to record where the data came from.
- Storage limitation: you must not keep personal data for longer than you need it.
- Suppression: practitioner retention guidance says to keep a minimised record on a suppression file until you remove the risk of getting the individual's contact details again.
No published rule deals specifically with contacts whose employer has dissolved, so agree the handling with your privacy owner and write it into the run notes. By outcome:
- Live and Renamed: no contact action from the sweep itself.
- Absorbed and Dissolved with a trading successor: re-point a contact only after confirming their current employer. Rectify their record with the new source and date.
- Defunct: suppress contacts who cannot be confirmed elsewhere.
- Domain Lost: stop sending to addresses at that domain. Mail to it may reach whoever registered it.
suppression_id: email_hash_or_address: reason: employer Defunct | domain re-registered | unconfirmed after Absorbed source_account_id: evidence_row_id: sweep_date: recorded_by:
Keep only these fields; add a column only if your privacy owner can name the purpose it serves.
How do you reconcile the run and keep the sweep current?
The sweep is done when a second operator, starting from the frozen export and the evidence log, gets the same count for each outcome. If the counts differ, the gap is the exceptions log you still owe, not a rounding error.
Keep the evidence log in a form someone else can replay. Each row should be enough for a stranger to repeat the lookup and get the same answer on the same snapshot.
account_id,legal_name,registry_number,evidence_type,source,snapshot_date,finding,proposed_outcome,reviewer,final_outcome,exception_reason,recheck_date 0011,Example Ltd,01234567,registry,Companies House dissolved search,YYYY-MM-DD,dissolved,Defunct,AB,Defunct,, 0011,Example Ltd,01234567,domain,RDAP,YYYY-MM-DD,creation date after account created,Defunct,AB,Defunct,, 0042,Sample Corp,,lei,GLEIF record,YYYY-MM-DD,LAPSED,Live,AB,,LAPSED only; registry not matched,YYYY-MM-DD
One row per piece of evidence; an account can have several rows but only one outcome.
Report the before-and-after table as the run's headline: the account count at export, the count per outcome, and the number left in exceptions. Use that as your decay figure, not the untraceable 22.5%. It is the only rate you can defend to an auditor, because you can show every row behind it.
To keep the work current:
- Run the entity sweep quarterly. That matches the 90-day cadence vendor content recommends, and it is enough for registry changes.
- Run the domain check more often than the entity sweep. The drop window is about 35 days, so a domain-only RDAP pass between quarters catches re-registrations before they collect leads. This is my judgement from the windows above, not a published standard.
- Re-check Ohio Defunct outcomes while they are inside the two-year reinstatement window.
- Re-check every LAPSED-only exception at the next run. Do not promote it to Defunct by default.
Before you call the sweep done
- The frozen export has a logged date and row count, and it has not been edited since.
- Every non-Live account has at least one evidence row with a source and snapshot date.
- Every Absorbed account names a successor whose own record was fetched, not just read from a SuccessorEntity field.
- No outcome rests on a LAPSED LEI, a missing bulk-file row, or serverHold alone.
- Every domain has an RDAP creation date compared against the account's created date.
- No dead record still holds a domain as primary that a successor or new owner also uses.
- Every record that was merged was exported first, and the export is stored with the run.
- Contacts at Defunct and Domain Lost accounts are suppressed with a minimised record.
- A second operator re-ran the sweep from the frozen export and matched the per-outcome counts.
- Every exception has a reason and a re-check date.
Questions practitioners ask
How do I find out-of-business companies in my CRM in bulk?
Match each account by registry number, not by name, against the registry that holds it, then check the domain. In the UK, the free bulk file lists only live companies, so query the Companies House dissolved-search API for anything missing from it. In Ohio, use the weekly cancelled/dissolved list. Then pull RDAP status and redirect targets for every domain, because a company can stay legally registered while its domain drops and starts routing leads to the wrong record.
Does a lapsed LEI mean the company has closed?
No. LAPSED means the LEI was not renewed by its renewal date and the entity is not known from public sources to have stopped operating. The issuer checks that the entity is still operating before it sets LAPSED, so the status points slightly toward Live, not toward Defunct. Use RETIRED, a filled SuccessorEntity field, and the registry status for closure evidence, and treat a lapse as nothing more than an unpaid renewal.
Should I merge an acquired company's account into the acquirer or link them?
Link by default and merge only when you have exported the dead record and nothing unique on it is worth keeping. A Salesforce merge moves opportunities, contacts, activity, campaign history, and cases to the surviving account. It keeps the merged-away record in the Recycle Bin for only 15 days, documents no unmerge, and loses unique field data you did not choose to keep. A link keeps the history auditable and can be undone.
Why do new leads keep attaching to an account for a company that no longer exists?
Usually the old domain is still on the record. Either it dropped and someone else registered it, so their form fills match your dead account, or two HubSpot company records share it, and HubSpot then sends new contacts with that domain to only one of them without letting you choose. Strip the domain from the dead record, and compare the domain's RDAP creation date with the account's created date to catch re-registrations.
What annual B2B data decay rate should I plan for?
No published rate holds up. The widely repeated 22.5% figure traces back to a study nobody has found, and it is about a decade old. Vendor claims of 30 to 40% a year at startups do not publish their methods. The defensible number is your own: the before-and-after count by outcome from each sweep, measured against the same frozen export, quarter after quarter.
Can I keep contact records for people whose employer dissolved?
Only if you rectify them, and otherwise you should suppress them. UK guidance expects inaccurate personal data to be erased or rectified without delay, with its source recorded, and data not kept longer than needed. Practitioner guidance recommends keeping a minimised record on a suppression file so the person is not re-imported. No published rule deals specifically with contacts whose employer has dissolved, so take the decision with your privacy owner.
Try it on the search you came here for
Stop building boolean strings. Just describe the person.
Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.
01Describe them
One plain sentence. Role, city, stack, stage, whatever matters to you.
02I read the web live
GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.
03You read the shortlist
Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.
- Staff backend engineers in NYC who shipped Rust in production
- Series A fintechs in SF under 50 people, growing headcount this year
- Maintainers of fast-growing Rust web frameworks on GitHub
- No boolean, no filters, no seat to buy. One box.
- Read at search time, so a profile updated yesterday counts today.
- Every step visible as it runs, every name with its reason.
500 free credits on sign-up. No card, no demo call. See real searches.