You bought a list eight months ago, or built one, and you want to know how much of it is still worth sending to. The number you need is about your list. Almost everything published is about somebody else's.
The good news is that three of the four things people mean by decay are measurable on your own data in an afternoon, and the fourth is measurable over a quarter.
Four things, one word
Before any number means anything, decide which quantity you are asking about.
An address can be dead: the mailbox no longer exists and mail to it hard bounces. A person can have moved: the address still works, someone still reads it, and the person you researched is at a different company. A contact can have gone quiet: the address works, the person is still there, and they have not opened anything from you in a year. Or a record can be stale in one field: the email is fine and the job title, company name or phone number attached to it is out of date.
These are four different quantities. They do not move together, they do not have the same cost, and the fixes are different. A hard bounce threatens your sending reputation this week. A job change makes your message land wrong but costs you nothing technically. Published decay figures routinely collapse all four into one percentage, which is why two sources can both be honest and disagree by a factor of three.
Measuring each one on your own list
Dead addresses. Take a random sample from the list, not the most recent records and not the ones you like. Send them something real and low-stakes, and count hard bounces against messages delivered. That gives you an invalid rate you can apply to the rest without having risked the rest. On our own lists a few hundred records is the sample we start from, which is a working rule rather than a statistical result. One class of address will not answer this test: a domain configured to accept every address returns no bounce whether or not the mailbox exists, so those records stay unresolved and belong in a bucket of their own rather than in your invalid rate.
Job changes. Take a smaller sample, a few dozen records, and check the current employer on each. This is the slow one, and it is the one that most affects whether a message works, because a correct address attached to a wrong company produces a reply you cannot use. Do it by hand once and you will know whether it is worth automating.
Quiet contacts. If you have mailed this list before, your platform already has the answer: segment by last open or last click and count what has not engaged in six months, twelve, twenty-four. If you bought the list and have never sent to it, there is nothing to measure here yet, and your first real send is what creates the baseline. Do not let a vendor's engagement figure stand in for one you have not generated.
Stale fields. Use the same small sample and check title and company against the person's current public profile. Count a record as stale if any field you personally use for targeting is wrong.
Four numbers, on your list, with your acquisition source and your sector baked in. Now the published figures become useful as context rather than as substitutes.
What the published figures are measuring
The most complete public series comes from ZeroBounce, whose Email List Decay Report gives a year-by-year figure: 23% in 2021, 22% in 2022, 25% in 2023, 28% in 2024 and 23% in 2025.
The denominator is stated on the page, which is more than most sources manage. In 2025 the company's software processed more than 11 billion email addresses, in bulk and in real time, and the report covers everything it verified between January and December of that year.
Two things about that population matter before you apply the number to yourself. It is addresses that somebody chose to submit to a verification service, which is not a random sample of business contact lists. People send lists to a verifier when they suspect there is a problem. And the customer base is described on the same page as ranging from solo business owners to Amazon, Disney and Netflix, so it is not a B2B measurement either. The report also carries no visible publication date.
ZeroBounce sells email verification, so the report is published by the company whose product fixes the problem it describes. That does not make the figure wrong. It makes it a figure with a stake in the answer, measured on a self-selected population, and worth reading alongside your own four numbers rather than instead of them.
Where the most-repeated number comes from
One figure appears on nearly every page about this subject: 2.1% a month, annualised to 22.5% a year. It is usually attributed to HubSpot, sometimes to MarketingSherpa.
HubSpot's page does exist and does state it. The wording is that email marketing databases naturally degrade by about 22.5% every year, footnoted to MarketingSherpa's research showing B2B data decaying at 2.1% per month. The footnote has no link, no report title, no year and no method. Every other statistic on that same HubSpot page is sourced to HubSpot's own State of Inbound reports from 2013 and 2014, and the page carries no publication or last-updated date of its own.
Following it back one more step does not resolve it. The figure does not appear on MarketingSherpa's own site, and it is not in the 2011 B2B Marketing Advanced Practices Handbook, the only MarketingSherpa primary document that surfaced. No page citing the number anywhere names which report it came from.
It is still in circulation as current, republished as recently as April 2026 as a present-tense fact about B2B lists.
Use it as an order of magnitude if you like. Do not use it as a benchmark you are meeting or missing, because there is nothing behind it to compare yourself to.
While you are reading the category, one more pattern is worth recognising. Several pages publish decay percentages broken down by segment, giving separate figures for free SaaS lists, purchased lists, healthcare and government contacts, and cold outreach lists. Producing numbers at that granularity would require tracking a labelled panel of contacts over time by acquisition channel and industry. No vendor claims to run one and none of those pages cites a method. Treat that level of detail as a signal about the page rather than about your list.
A number that is genuinely measured
For the job-change component there is a real, independent, monthly measurement, and it is free.
The US Bureau of Labor Statistics publishes JOLTS, which surveys employers directly. For the June 2026 reference month, released on 4 August 2026, the quits rate was 2.0% and the total separations rate, which adds layoffs, discharges and other separations, was 3.4%.
Those are monthly rates against total employment, and the temptation is to multiply. Do not. A monthly separations rate and an annual share of contacts who have moved are different quantities, and BLS publishes nothing that converts one into the other. Twelve times 2.0% is arithmetic performed on a number that was not built to carry it.
What the JOLTS number is good for is direction and floor. Roughly one in fifty US workers leaves a job in a typical month, which tells you that a list untouched for a year has meaningfully moved and gives you a sourced, dated, non-commercial reason to say so.
LinkedIn's Work Change Report, published January 2025, is the other name that comes up here. Its stated finding is that people entering the workforce now are on pace to hold twice as many jobs over a career as those who started fifteen years ago. That is a career-long cross-cohort comparison and it cannot be converted into an annual rate.
What decay costs before it costs you money
The first cost is not wasted messages. It is your sending reputation.
Google's bulk sender requirements, which have applied to senders of 5,000 or more messages a day since 1 February 2024, tell senders to keep spam rates reported in Postmaster Tools below 0.30%, and separately recommend staying below 0.10% and never reaching 0.30%. That is a spam complaint rate rather than a bounce rate, and Google publishes no numeric bounce ceiling. The connection to a stale list is indirect and real: mail to people who no longer recognise you attracts complaints, and complaints are the metric with a published number attached.
On bounces the record is thinner than you would expect. Mailchimp's own documentation confirms that account suspensions are automatic actions triggered when a send exceeds industry thresholds, and does not publish what those thresholds are. The commonly quoted 2% bounce ceiling was not traceable to any provider's own documentation in this research. It is traceable to a data vendor publishing it as the standard to expect from any paid B2B provider including itself, which is a different kind of source from a platform rule and worth weighing as one.
Setting a refresh cadence from your own numbers
Run the four measurements once and you have a baseline. Run them again a quarter later on a fresh sample and you have your rate, on your list, in your sector.
From there the cadence is arithmetic rather than judgement. If your invalid rate is climbing toward the point where a full send would risk your spam rate, verify before the next campaign. If your job-change rate on a hand-checked sample is high, the list needs re-researching rather than re-verifying, and those are different purchases. If quiet contacts dominate and the addresses are fine, the problem is the message rather than the data. If you are hoping to hand that re-researching to a model, what Claude reaches and what it does not is worth reading before you build the workflow.
The figures in this article are worth about one paragraph of your planning. The four you measure yourself are worth the rest.
Frequently asked
- What is email list decay?
- The gradual loss of usable contacts from a list over time. The term covers four different things that are often reported as one number: addresses that no longer exist and hard bounce, people who have changed employer, contacts who have stopped engaging, and records where a field such as job title or company is out of date. They move at different speeds and need different fixes.
- What is the average email list decay rate?
- The most complete public series comes from an email verification company and reports between 22% and 28% a year across 2021 to 2025, measured on addresses its customers submitted for verification. That population is self-selected, since people send lists to a verifier when they suspect a problem, and it includes consumer as well as business senders. The monthly figure that most pages repeat has no locatable original study behind it.
- How do I measure decay on my own list?
- Take a random sample and send it something real, then count hard bounces against delivered messages for your invalid rate. Take a smaller sample and check each person's current employer by hand for your job change rate. Sample sizes are your judgement; a few hundred for the send test and a few dozen for the hand check is a working starting point rather than a statistical rule. Your sending platform reports engagement by segment. Check title and company on the same small sample for field staleness.
- Does the BLS quits rate tell me how fast my list decays?
- It tells you the rate at which US workers leave jobs, which is one driver of decay rather than a measure of it. For June 2026 the quits rate was 2.0% and total separations 3.4%, both monthly rates against total employment. A monthly separations rate and an annual share of contacts who have moved are different quantities, and nothing BLS publishes converts one into the other.
- How often should I clean a B2B list?
- Often enough that a full send does not put your spam rate at risk. Google asks bulk senders to keep the spam rate reported in Postmaster Tools below 0.30% and recommends staying under 0.10%. Rather than following a fixed schedule, measure your own invalid rate on a sample before a large campaign and verify when it has moved enough to matter.