Firmographic data stays accurate when each field is refreshed at the speed it changes: company status and headcount every month, headquarters address and website every quarter, revenue with each new filing and industry code once a year and at each NAICS revision. One cycle for every field leaves the fast fields stale between refreshes.
This post sets out the official change rates behind those cycles, taken from the Census Bureau's Business Dynamics Statistics and the BLS Business Employment Dynamics release for the first quarter of 2026, both read on 15 November 2026. It then gives a field by field table, a worked example and a decay test you can run on your own CRM.
How often should firmographic data be refreshed to stay accurate?
Firmographic data should be refreshed monthly for the fields that say whether a company still trades and how large it is, and quarterly to yearly for the rest. The Census Bureau put the US establishment exit rate at 9.4 percent and the job destruction rate at 11.1 percent for the year to March 2023.
Both figures come from the economy-wide table of the 2023 Business Dynamics Statistics, the latest BDS release, which measures each year from one March to the next. An exit rate of 9.4 percent a year works out to about 0.8 percent of establishments a month, which is roughly the share of accounts that can close between one monthly refresh and the next.
A rule of thumb follows from that. The months between refreshes of a field equal the share of wrong records you will accept divided by the field's monthly change rate. If you accept 2 percent of accounts showing as active after they have closed, 2 divided by 0.8 gives two and a half months, so company status needs a monthly cycle in practice. A field that changes 2 percent a year passes the same test with a yearly refresh.
Scheduled refreshes work best with change events between them. A registry status change, an acquisition announcement or a new filing should update the record when it arrives, so the next scheduled cycle only has to catch what the events missed.
How fast does each firmographic field go stale?
Company status, headcount and the list of companies in your market all change every quarter in the official data. The table gives the official measure we read for each field and the refresh cycle it supports. Where no official series was found, the decay test later in this post fills the gap.
| Field | What changes it | Official measure we read | Suggested refresh |
|---|---|---|---|
| Company status (active or closed) | Closures and dissolutions | BDS: 523,958 firm deaths in the year to March 2023, against 5,593,727 firms in March 2023 | Monthly, and on each registry change |
| Companies in your market | Openings and new firms | BDS: establishment entry rate of 10.6 percent in the year to March 2023; BED: 323,000 establishment births in the first quarter of 2026 | Monthly |
| Headcount and size band | Hiring, lay-offs and closures | BED: gross job gains and gross job losses each 5.5 percent of private-sector employment in the first quarter of 2026 | Monthly for scoring, quarterly for segments |
| Revenue | A new financial year | We found no official series on revenue changes at company level | Yearly, and on each new filing |
| Industry code | A change of main activity, or a revision of the code list | NAICS is reviewed every five years; a 2027 revision is proposed for data on periods from 1 January 2027 | Yearly, and at each revision |
| Headquarters address | Moves and mergers | We found no official US series on business moves | Quarterly, checked with the decay test |
| Legal name and parent | Rebrands, mergers and acquisitions | GLEIF publishes new and changed LEI records three times a day | Monthly, and on each deal announcement |
| Website domain | Rebrands, acquisitions and site migrations | We found no official series | Quarterly |
The NAICS row comes from OMB's notice of 13 July 2026, which states that NAICS is reviewed every five years and that the 2027 revision will be the sixth since OMB adopted NAICS in 1997. NAICS United States 2022 has 1,012 national industries, so a company whose main activity shifts can move to a new six-digit code without any change to the code list itself.
Two limits apply to every row. The BDS and BED count establishments and jobs across the US economy as a whole, so their rates are averages over businesses of every size, and a CRM weighted towards larger accounts may see a different rate. They also say nothing about how often a company moves from one headcount band to another, which is the change most scoring models use.
An establishment and a firm also differ. A company that closes one of its sites stays active, so for account status the firm death count is the closer measure: 523,958 deaths set against 5,593,727 firms is about 9.4 percent on our arithmetic. The BDS frequently asked questions add that a firm whose legal entity ends through a merger or acquisition is not counted as a firm death, so the count leaves out accounts that change hands. The legal name and parent rows need their own check for that reason.
What is the decay rate of B2B firmographic data?
The primary sources checked for this post publish no single decay rate for all B2B firmographic data, so a single annual figure is a claim to test. The official series measure each kind of change separately, and the refresh frequency for a field follows from its own rate and the error you can accept.
Decay in a CRM is the share of records with at least one field that no longer matches the company. Because fields change for different reasons, that share grows faster than the rate of any one field. If three fields change at 9.4, 15 and 5 percent a year and the changes are independent, about 27 percent of records carry at least one stale field after twelve months, as the worked example below shows.
Headcount moves faster than annual figures suggest. In the BED release published on 28 October 2026, private-sector establishments gained 7.4 million jobs and lost 7.3 million between December 2025 and March 2026, each 5.5 percent of private-sector employment. The net change was 88,000 jobs, so a flat total hid gross flows of more than 7 million jobs in each direction within one quarter.
Official counts of closures also arrive late. BLS counts an establishment as a death only after four consecutive quarters with no employment, so the October 2026 release carries deaths only to the second quarter of 2025, when 923,000 jobs were lost at 310,000 establishments. For single accounts, the registry entry, the company's website and its announcements give earlier evidence than the statistics.
A worked example
This illustrative example shows how the refresh cycle changes the number of stale records in a CRM. Only the closure rate comes from an official source.
Assumptions: Acme Robotics keeps 10,000 US accounts in its CRM, all verified on the same day. Accounts close at the BDS establishment exit rate of 9.4 percent a year. In a year, 15 percent of accounts cross a headcount band and 5 percent move headquarters; both figures are invented for illustration. Each annual rate is converted to a steady compound monthly rate, and the three fields change independently.
| Refresh cycle | Closed accounts still marked active, just before refresh | Accounts with at least one stale field, just before refresh |
|---|---|---|
| Monthly | 82 | 257 |
| Quarterly | 244 | 752 |
| Yearly | 940 | 2,684 |
At 9.4 percent a year, the monthly closure rate is about 0.82 percent. The share of records with no stale field after a year is 0.906 times 0.85 times 0.95, or about 0.73, so about 27 percent carry at least one stale field by the time a yearly refresh comes round.
Under these assumptions, a yearly cycle leaves 2,684 of 10,000 records with a stale field just before refresh, and a monthly cycle leaves 257, about one in forty. The invented headcount rate drives most of that gap, so the next step is to measure it on your own accounts. Your rates will also differ by size band, country and industry, so read the table as a comparison of cycles.
Should a provider refresh firmographic data daily, weekly or monthly?
A provider should refresh firmographic data daily or weekly for fields that trigger action, such as company status, leadership and headcount, and at least monthly for slower fields such as industry code. Check the date each value was last observed, because a file delivered daily can still carry values observed weeks earlier.
The cadence you need follows the decision the data feeds:
- Daily, for alerts and triggers, where an action follows a change within days: a closure, an acquisition or a new location.
- Weekly, for account scoring, where scores are recalculated on a fixed day and compared week on week.
- Monthly, for segmentation, territory planning and market sizing, where counts across many companies matter more than any single change.
Three intervals sit behind the word refresh: when the provider observed the value, when it delivered the file and when you loaded it. The guide to technographic data refresh frequency sets out that split for technology data, and it applies to firmographic fields in the same way.
Your own load interval adds to the delay. If you join provider data to CRM tables in a warehouse, write each match to a dated table, as the Snowflake join guide recommends, so you can tell later which version a report used.
How often do B2B data providers update firmographic data?
B2B data providers can update a field no faster than the source they take it from, and the official sources run on different cadences. GLEIF publishes its LEI files three times a day, while Companies House publishes its free company snapshot once a month, so ask each provider which source and date sit behind every field.
The GLEIF Golden Copy files come out at 02:00, 10:00 and 18:00 UTC, each with delta files that list new and changed LEI records against the files of eight hours, 24 hours, seven days and 31 days earlier. The UK Free Company Data Product is compiled to the end of the previous month and updated within five working days of month end; on 15 November 2026 the latest file listed was dated 1 November 2026. A provider that loads that snapshot can be more than a month behind on a UK company's registered details, whatever its delivery schedule.
Fokals states its cadence per source on its coverage page. Company websites are read daily for listed companies in priority markets, every three days for other listed companies and weekly for private companies. Company hierarchy is updated monthly and with each new annual report, and business verification runs monthly, with a weekly run for US filers. The delivery page sets out the cadence of each dataset, from daily to monthly, and delivery by REST API or as bulk files.
Before you rely on any provider's stated cadence, Fokals included, ask for these points in writing:
- The date that travels with each field: when it was observed, when it changed or when it was delivered.
- The source behind each field and how often that source publishes.
- Whether a change arrives as a dated event or overwrites the old value.
- How a closed company is flagged and how soon after the registry records the closure.
- Whether point-in-time history is available, so a model can be tested on what was known at the time.
- A match against a sample of your accounts, so you can run the decay test below on the provider's records.
A decay test you can run on your own CRM
You can measure the decay rate of your own firmographic data with a random sample of about 200 accounts, a primary source for each field and one query. The result is a monthly change rate per field, which sets that field's refresh cycle more reliably than any borrowed average.
- Draw 200 accounts at random from records last verified at least six months ago, and keep each record's last verified date.
- Pick the fields to test: company status, legal name, headquarters address, website domain, headcount band and industry code.
- Check each field against a primary source: the company registry for status and legal name, the company's own website for address and domain, and a dated filing or announcement for headcount.
- Mark each field as changed, unchanged or unverified, leave the unverified group out of the rates and report its size.
- Compute a monthly rate per field and set the cycle with the rule of thumb above: months between refreshes equal the error you accept divided by the monthly rate.
- Repeat on a fresh sample each quarter, split by size band and country.
With the results in one table, one row per account and field, a query in Snowflake syntax gives the rates:
select field,
count_if(result = 'changed') as changed,
count_if(result in ('changed', 'unchanged')) as verified,
avg(months_since_verified) as avg_months,
changed / nullif(verified, 0) / avg_months as monthly_rate
from decay_audit
group by field
order by monthly_rate desc;With 200 records, an observed rate of 10 percent carries a 95 percent margin of error of about 4 percentage points either way, so treat small gaps between fields as noise. Over gaps longer than a year the simple division understates the monthly rate a little, because changes compound.
The test measures change since the last verification. It cannot catch a record that was wrong from the start, such as an account matched to the wrong company, so run a match-rate check alongside it.
Frequently asked questions
Why do official closure figures arrive later than figures on openings?
BLS counts an establishment as a death only after it reports no employment in four consecutive quarters, so a closure can be confirmed only about a year after the last quarter with staff. That is why the October 2026 BED release carries births to the first quarter of 2026 and deaths only to the second quarter of 2025. Use the statistics to set refresh cycles, and the registry, the website and announcements to catch single closures.
Does a merger count as a closure in the official data?
In the Census Bureau's Business Dynamics Statistics, a firm whose legal entity ends through a merger or acquisition is not counted as a firm death. In a CRM, an acquired account may need a new parent, a new legal name or a new domain, so track deals as their own event type and update those fields when a deal completes.
Should headcount and revenue be refreshed on the same cycle?
They move at different speeds. In the BED release for the first quarter of 2026, gross job gains and gross job losses each equalled 5.5 percent of private-sector employment in three months, while revenue moves with the financial year and the filings that report it. Refresh headcount monthly, update revenue when a new figure is published and store the date beside each value so a reader can tell how old it is.
What should I do with stored industry codes when NAICS 2027 takes effect?
OMB's notice of 13 July 2026 proposed a revision that would apply to establishment data for periods beginning on or after 1 January 2027, and it lists how the proposed industries link to NAICS 2022. Keep the code version beside every stored code, re-map records once the final changes are published in the Federal Register and keep the old code for history.
Does refreshing fix records that were wrong from the start?
A refresh catches change since the last check. An account matched to the wrong company stays wrong however often it is refreshed, so measure the match rate separately: divide matched accounts by accounts with a usable domain and review the misses by cause.
The queries and code on this page are examples to adapt. Test them in your own environment before you rely on them.
What this page says about the products it names was checked against their public documentation on 15 November 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.