In short
- B2B contact data decays somewhere between 22% and 30% a year, depending on whose method you use.
- A widely repeated 70% figure traces through vendor citations to no published method. Do not use it.
- Order of operations matters more than tooling: deduplicate, verify, consent-check, suppress, segment.
- Consent status is usually checked last and is the step most likely to stop a campaign.
- Cleaning fields nobody uses is theatre. Clean what drives routing and personalisation.
- The measurable output is reachable contacts before and after, which is a number the business can quote.
Why this job is never finished
Contact data goes out of date on its own. People change job, companies restructure, domains change after an acquisition, and nobody tells your CRM.
The commonly cited annual decay rate for B2B contact data is around 22.5%, from a methodology that has been reproduced across the industry for years. ZoomInfo’s own research puts it higher, at 25 to 30%. Both are vendor figures, and both are close enough to be useful as a planning assumption: a database left alone for a year has lost somewhere between a fifth and a third of its accuracy.
You will also find a figure of 70.3% a year in circulation. Follow it back and it traces through a chain of vendor citations without arriving at a published method. It is not usable, and the fact that it keeps appearing tells you how much of this field runs on repetition rather than measurement.
The order matters more than the tool
Most articles on this subject recommend software. The sequence matters more, because doing these steps out of order means paying to clean records you were about to delete.
1. Deduplicate first
Duplicates inflate every number downstream. Contact counts, cost per contact, engagement rates, and the size of every segment you are about to build.
Vendor analyses put B2B duplicate rates somewhere between 10 and 30% of records. Whatever yours is, finding out costs an afternoon and changes every figure you report afterwards.
Deduplicate before verification, because verifying two copies of the same person costs twice as much and produces one useful answer.
2. Verify what remains
Email verification against the deduplicated set. This removes the hard bounces before they damage a send, which matters most on a list that has been quiet, since bounces on a cold list are read harshly by mailbox providers.
3. Consent-check
This is the step that gets done last and should be done third, because it can shrink the list dramatically and everything after it depends on the result.
For every contact, what is the lawful basis, and where is it recorded. Contacts without one go to suppression rather than into a sequence. This is not a technicality that can be tidied later. It decides who you are permitted to mail.
4. Suppress rather than delete
Contacts you cannot verify, cannot consent-check, or who have complained go to a suppression list. They stay as records, so the audit trail survives and nobody re-imports them next year from an old spreadsheet.
Deleting destroys the evidence that you removed them for a reason.
5. Segment last
Only now is segmentation worth doing, because only now do the segment sizes mean anything.
Clean the fields that do something
A full field-by-field cleanup of a CRM is a project that never ends and rarely pays.
Clean the fields that drive an action:
- Fields used for routing. Country, company size, product interest. A wrong value here sends a lead to the wrong person or to nobody.
- Fields used in personalisation. First name and company name, because these appear in the message and errors are visible to the recipient.
- Lifecycle stage and lead source. Because reporting and automation both key off them.
Everything else can wait, and most of it can wait forever.
The number to produce at the end
The output of this work is one figure the business has probably never seen: reachable contacts.
Total records, minus duplicates, minus unverifiable addresses, minus contacts with no lawful basis, minus suppressions. What remains is the audience you actually have.
That number is almost always smaller than the one being quoted in board meetings, and it is the only one worth planning against. It also produces the honest version of the value argument, because the difference between the two figures is the size of the illusion the business had been working with.
Questions, answered
How often should this run? Deduplication and verification quarterly, consent-checking whenever the process for collecting consent changes, and a full pass annually. The decay rate makes anything less than annual a losing race.
Should we buy a data enrichment tool? Only after the clean. Enriching a database full of duplicates and unverifiable records means paying to improve data you are about to remove.
Can we clean the list by just sending to it? This is a common shortcut and it damages the domain. Bounces and complaints from an uncleaned list are precisely the signals mailbox providers use to decide whether to accept your mail at all.
What about contacts who never opted in but came from an event? Depends on the jurisdiction and how the event captured them. The safe route is to treat missing evidence as missing consent, and to fix the capture process so it stops happening.
How long does a first clean take? For a database of twenty thousand contacts, a fortnight is realistic, and most of that is waiting on verification and on internal answers about where consent was recorded.
Sources
- B2B data decay of roughly 22.5% a year: MarketingSherpa methodology, widely reproduced. Vendor data.
- 25 to 30% annual decay: ZoomInfo research. Vendor data.
- Duplicate rates of 10 to 30%: vendor analyses, 2026. Vendor data.
- The 70.3% figure: traced 2026-08-04 through vendor citations to no published method. Listed here as unusable.
This is one part of a bigger picture. Start with our guide to B2B email marketing for the whole sequence.