The Field Decay Reference: What Goes Stale and How Often to Re-Verify

This is a field-by-field lookup for anyone answerable for how a record was gathered and how long it can be trusted: RevOps and revenue operations owners setting refresh policy, and the data owners who defend it. It exists because public answers repeat one blended annual number that hides how differently email, phone, title, and firmographic fields age. Jump to the row you need, read the decay rate, the trigger events that kill it, how it fails silently versus loudly, and the cadence it earns, and set policy without wading through a sales pitch.

Why one blended decay number misleads you

The single most-cited figure, 22.5% per year, is a contact-record rate, not a field rate. MarketingSherpa measured B2B data decaying at 2.1% per month, and HubSpot's Database Decay Simulation compounds that to about 22.5% a year. That number is real and useful for sizing a whole database. It is wrong the moment you use it to budget refresh effort per field.

The reason is that fields on the same record age at wildly different speeds. Job titles change roughly 65.8% a year. Founded year changes never. If you set one refresh interval for both, you burn cycles re-checking dates of incorporation while a third of your titles quietly rot. A single blended rate averages a field that turns over three times a year with one that never moves, and the average describes neither.

Two cautions on the numbers themselves. A widely repeated "70.3% a year" figure attributed to Gartner does not trace back to any locatable Gartner report, so treat it as folklore. And precise per-field monthly rates for phone and address are largely vendor-asserted rather than published with methodology - useful as direction, not as measured fact. This reference marks which figures are which so you know how hard to lean on each row.

The decay rate table: what each field is worth per year

Here is the spine. Every rate below carries a named source, and the two fastest fields - title and email - are the ones outreach depends on most, which is the structural problem this whole reference exists to manage.

Read this as a priority order, not a schedule. Titles at 65.8% and firmographics at 30% dwarf the aggregate, while email sits close to it. A harsher counter-benchmark exists: ZoomInfo's analysis puts contact decay at 25 to 30% a year, above the classic 22.5%. Where two credible sources disagree, plan to the higher rate and let your own instrumented bounce and connect rates settle the argument.

What each rate proves and how it lies:

• Title (65.8%) proves how often someone's role label stops matching reality. It lies by looking stable when a promotion within the same company changes seniority but not employer, which your enrichment may miss entirely.
• Email (23%) proves list-level degradation. It lies through catch-all domains that verify green while the mailbox may not exist - covered in its own section below.
• Phone (18%) proves number churn. It lies quietly: a reassigned or disconnected number does not always self-report until you dial it.
• Firmographic (~30%) proves company-level drift. It lies most on revenue, which lags real numbers by two to four quarters, and on headcount, which drifts silently between imports.

Trigger events: what invalidates a field and how fast

Decay is bursty, not smooth, because most of it is triggered by discrete events rather than steady erosion. A job change is the dominant trigger, and it hits several fields at once. When someone gets promoted, moves to a competitor, or leaves the workforce, every field tied to their old role goes stale together: title, work email, and direct-dial number, all invalidated by one event.

That is why field-independent cadences are inefficient. They over-refresh the people who never moved and under-catch the ones who did. About 30% of professionals change jobs annually, and each change writes off three or four fields simultaneously across every database holding them.

The timing matters for cadence. A departing employee's company email typically deactivates within 30 to 90 days, which sets the window you have to catch a mover before the old address stops verifying. Company events run on their own clocks: domains change on a rebrand, mergers fold two records into one while many vendors keep both alive for months, and revenue estimates lag two to four quarters behind reality.

Silent versus loud staleness

A field fails loudly when it self-reports, and silently when your record simply stops matching reality with no error. Loud failures are the easy ones: a hard bounce or a disconnected number tells you the field is dead. Silent failures are where the real cost sits, because none of it shows up as an error message - your records just quietly stop matching reality.

The most expensive silent failure is the catch-all email. Catch-all domains accept every address sent to them even when the specific mailbox does not exist, so they are impossible to validate without emailing and waiting to see whether they bounce. Verification returns green, the send goes out, and non-existent addresses bounce afterward, damaging sender reputation. Catch-alls were more than 9% of all checked emails in a recent report - a large, invisible slice of any list.

To make silence measurable, instrument field-level metrics rather than reading aggregates. Two carry the most weight:

• Percent verified within 90 days. Keep this above 80% on revenue-driving fields. It is the single freshness metric that maps directly to cadence.
• Connect rate. The percentage of calls that reach a live person. A well-maintained database benchmarks at 15 to 25%; below 10% signals widespread phone decay.

Score databases on accuracy, completeness, freshness, validity, and consistency, and weight accuracy and freshness most heavily, because a record can be complete and consistent yet wrong. Watching one global bounce rate hides the damage: a healthy transactional stream at 0.1% can mask a cart-recovery flow at 3.2% in the same aggregate. Break every metric out by segment and source.

Refresh cadence: the calendar floor and the event layer

The published consensus is a quarterly bulk floor layered with event triggers, and the reason trigger beats calendar is that decay is not evenly distributed. A 22.5% annual baseline suggests quarterly bulk checks at minimum, with trigger-based refreshes on job-change or engagement signals for high-value accounts on top.

Differentiate by class. Verify contact fields far more often than firmographics, and inside firmographics, split high-decay from low-decay attributes: employee count and growth rate need quarterly attention, while founded year never needs refreshing.

There is genuine disagreement at the fast end. Some vendors argue for monthly cleansing on large email lists (50,000+), while the published floor is quarterly. Resolve it with your own numbers: if bounce rate climbs past 1 to 2% inside a quarter, tighten toward monthly for that segment. Recency beats size everywhere - a source of 30 million profiles checked every quarter is worth more than 200 million on an 18-month re-check cycle, because at 2% per month freshness compounds faster than coverage.

The hardest part is catching movers inside the 30-to-90-day email window rather than at the next quarterly batch. Subscribing high-value records to change signals is what closes that gap, and it is where a signal-driven search earns its place: instead of waiting for a bounce, you pull the people who just changed roles and refresh their records before the old address dies.

Refolk lets you ask for those job-change cohorts in plain English rather than maintaining brittle trigger rules, which turns the event layer from a data-engineering project into a query. When a market looks empty, the same plain-English approach lets you check whether the field is genuinely stale or whether your filter simply missed it.

How much cadence a sector earns

Sector churn sets a floor on refresh frequency before any tooling choice, and job tenure is the cleanest public proxy for it. The BLS median tenure fell to 3.9 years in January 2024, the lowest since 2002, and the spread across industries is wide enough to change your cadence.

The tenure figures are BLS; the turnover proxy is a rough 1/tenure approximation, not a BLS number, so use it for relative ranking rather than as a decay rate. The signal is the ratio: leisure and hospitality churns at roughly twice the rate of manufacturing, which means contact records in high-churn verticals need roughly twice the refresh frequency of stable ones. Public-sector records, at 6.2 years median tenure, can tolerate a noticeably slower cadence.

One practitioner claim to treat as directional only: leadership and sales roles are said to churn nearly twice as fast as engineering, but that multiplier is asserted without published methodology. Weight your senior and sales fields more heavily than average, but do not encode "2x" as a measured constant.

The fields you most want are the ones that rot fastest, which is a structural mismatch between value and half-life.

The re-verification procedure

This is the eight-step sequence to turn the tables above into policy your team actually runs. It moves from inventory to instrumented, self-correcting cadence.

The one step teams skip is verification at the point of use. A quarterly batch keeps the database healthy on average, but the segment you are about to dial or send to is a small slice that may have drifted since the last cleanse. Re-verify that slice immediately before the touch, on top of the batch. It is cheap, it is targeted, and it catches the exact records where staleness would cost you a bounce or a wasted dial.

Where this goes wrong: failure modes and false positives

This is the most valuable section in the reference, because every failure below produces a green-looking record that is actually wrong. Each pairs the false positive with the local check that exposes it.

Trusting the blended number as a per-field rate. Budgeting equal refresh for email and founded-year is the default mistake. Check against the decay table: titles run about three times the aggregate, static fields near zero. Set cadence per field.

Catch-all emails scored as valid. They accept everything, so verification returns green while the mailbox may not exist. Segment catch-alls (about 9% of lists) and score them separately, watching actual bounce on send rather than verification status.

Calendar-only cadence. Because decay concentrates in events, a quarterly batch can miss a job-changer for up to 89 days while their old email still verifies. Reconcile batch dates against job-change signals and add an event layer.

Reading one aggregate bounce rate. A global rate of 0.8% feels safe, but a transactional stream at 0.1% can be masking a flow at 3.2% that is doing real damage. Break bounce out by segment and source, never global.

Company-name as the primary key. Rebrands and M&A silently break the join when the name changes. Key on domain, not name, and monitor domain-change events.

Match rate mistaken for accuracy. A match rate measures how often a vendor can find any data point; an accuracy rate measures how often that point is correct once found. Blind-test against known closed-won accounts to separate the two.

Title-string filters that miss non-English markets. A stale or empty field can be a query artifact, not a decay signal. Query localized titles before concluding a market is dead - the section below shows why.

Assuming firmographics are safe. Headcount and revenue drift silently between annual imports. Refresh employee count quarterly; never refresh founded year.

When a stale field is really a query artifact

Sometimes a field looks decayed when the real problem is normalization in how you queried it. Before you declare a segment dead and refresh it aggressively, confirm you are not looking at a filter that missed live people.

In Refolk's index of professional profiles, the current holders of "VP Sales / Vice President of Sales / VP of Sales" number 46,217 in the United States but only 242 in Germany for those identical English title strings.

The 191-to-1 gap is not a real absence of German sales leaders. It is mostly title-language normalization: German records less often carry English title strings, so an English-only filter reads as an empty market. The lesson for decay policy is that a field showing zero or near-zero coverage in a segment may be a normalization artifact, not a signal that everyone left. Query localized title variants before concluding the field decayed. There is also a caution the other way: the RevOps pool at 1,531 in the US is genuinely small, which is worth knowing when you rely on those people to own refresh policy at scale.

Keep the policy current

A field decay policy is not a one-time document; it drifts as your own data and your sources change. The way to keep it honest is to let your realized metrics correct your modeled rates every quarter.

Two disciplines keep this alive. First, at the quarterly review, compare the bounce and connect rates you actually observed against the decay rates you modeled; where they diverge, your data wins and the cadence table changes. Second, re-check the sources behind any figure you lean on hard, since email decay in particular has moved year to year (22% to 28% and back to 23% across recent reports), and the harsher 25 to 30% contact benchmark may become the consensus. Describe the mechanism, instrument the outcome, and let the numbers tell you when to tighten. Poor data quality carries a large enough cost - one widely cited estimate puts it at $12.9M a year for the average organization - that the review is cheap by comparison.