CRM data quality · a record is only as true as its last check
CRM data quality: how to keep CRM data true
The short answer
To keep CRM data quality high, stop treating cleanup as a yearly project. Records decay as people change jobs, get attached to the wrong account when matched on names or domains, and fill up with guesses. Match every record on a stable key, re-check it against a fresh source when something happens, and write back only what comes back found.
On this page11 sections
- What is CRM data quality?
- What causes poor CRM data quality?
- How fast does CRM data go out of date?
- How do you clean CRM data?
- How do you deduplicate CRM records without merging the wrong ones?
- Is CRM data cleansing the same as CRM data hygiene?
- How do you keep CRM data clean after the cleanup?
- What should you do when a source cannot confirm a record?
- How do you measure CRM data quality?
- How do you re-check a CRM record with the API?
- Sources and method
Most CRM records start out right. On the day a record is saved, the contact works where the record says, the account is the right one, and the email reaches a person. Then the world moves and the record does not. A few months later, reps call people who left, leads land on the wrong account, and the same company shows up twice in the pipeline report.
This guide explains where CRM data quality is lost, how fast it goes, and how revenue operations teams keep records true without a yearly cleanup project: one stable key per record, a re-check when something happens, and nothing written back that a source did not confirm.
What is CRM data quality?
CRM data quality is how far the records in your CRM still match the companies and people they describe, on the day someone uses them.
- Definition CRM data quality
- How far the records in your CRM still describe the real companies and people behind them: the right account, the job each person holds today, contact details that work, and one record for each.
PreciselyCRM data quality is the share of CRM records that match the company or person they describe on the day they are used. It fails three ways: records that were true when saved and have since decayed, records joined to the wrong account or duplicated because they were matched on a weak key, and fields filled with values nobody proved. It is kept by re-checking records against a fresh source, on events and on a schedule, matched on a stable identifier.
The definition has a date in it on purpose. A record is not good or bad once and for all: it was true when it was saved, and it stays true until something in the world changes. Quality is the distance between the two, and it grows on its own.
What causes poor CRM data quality?
Three faults, and each one hides behind a field that looks fine:
- Decayed
True when saved, wrong today
EmployerGlobex
She moved to Initech in March. The record never heard.
Re-check the current job - Mis-joined
Attached to the wrong account, or twice
AccountsAcme Inc · ACME Corp
Two accounts for one company, matched on the name as typed.
Match on one stable key - Unproven
Filled in with a guess
Emailfirst.last@acme.com
Built from a pattern. Nobody checked that the address exists.
Keep only what came back
- Decay. The record was true when it was saved. Then the contact changed jobs, got promoted, or the company moved or rebranded. This is data decay, and the CRM rarely notices it on its own: a bounce is often the first sign, and only a fresh outside check confirms it.
- Bad joins. The record is attached to the wrong account, or the same company exists twice. It happens when records are matched on what was typed: a company name written three ways, an email domain that is not the website, a subsidiary writing from its own domain.
- Guesses. A value was filled in without proof: an email built from a first.last pattern, a title inferred from a department, a company size estimated by hand. A guess looks exactly like a fact until it bounces.
Missing fields are the fault everyone sees first, and the least dangerous. An empty field tells you to look. A wrong value tells you to act, on the wrong thing.
An empty field tells you to look. A wrong value tells you to act, on the wrong thing.
How fast does CRM data go out of date?
As fast as your contacts change jobs, and the records most likely to be wrong at any moment are those of the people who just moved. Outside data sources fall behind on the same people.
BlitzAPI’s September 2026 benchmark shows where. The test looked up 3,445 people with six B2B data providers, queried on September 17 and 18, 2026, and checked each answer against the person’s public profile as read on the test day. Many of those people had changed jobs recently, on purpose: it is the situation that breaks contact data. Adding up the answers of the six providers (the report gives each one’s own figures):
- For people who had started their current job in the three months before the test, about one answer in three named a stale employer: a company other than the one the person worked for that day.
- For people who had held their job for more than a year, about one in 28.
- That is about nine times as often for recent movers as for people settled in their job.
Answers that named a stale employerBy how long the person had held their job on the test day
- 1 month or less785 people1 in 3
- 2 to 3 months916 people1 in 3
- 4 to 6 months255 people1 in 4
- 7 to 12 months448 people1 in 8
- More than 12 months1,041 people1 in 28
BlitzAPI's September 2026 benchmark: 3,445 people looked up with six B2B data providers, queried September 17 and 18, 2026. Answers added up across the providers; the report gives each one's figures.
The same answers say something useful about what a stale record is. In about nine cases out of ten, the stale employer was a company the person had really worked at, before their current job. Most of the time, the source was not inventing a career. It was late.
Read these numbers for what they are. The population was chosen to stress the data, so they are not a decay rate for your CRM, and a test measures every source against the profiles as they stood that day, including changes made just before it. What carries over to your database is the shape: a record goes wrong when its person moves, and the answers were stale far more often for people in their first months in a job than for people more than a year into it.
Two consequences for how you clean:
- The riskiest records in your CRM are the people who changed jobs recently, whether you know it or not. Re-check them first, and sooner.
- An enrichment that ran once is a snapshot. If it ran while the person was moving, it can have written the old employer back in. Look at the start date of the current job, not only at the employer.
Most stale records are not invented. They are late.
How do you clean CRM data?
Clean CRM data in a fixed order: settle the key first, then merge, then refresh, then remove what nobody can confirm. Done the other way round, a refresh writes fresh data onto the wrong account.
- Choose one key per object. For accounts, the company’s LinkedIn page and its numeric company ID. For people, their profile URL, as the API returns it. Both name an entity, where a company name or a domain only describes one.
- Backfill the key. Resolve each account’s domain to its company page, check that the page name matches the account, then read the company ID from the company profile and store it. Resolve each contact from their profile URL, or from their work email when that is all you have.
- Merge duplicates on the key (next section).
- Re-check every contact’s current job. Compare the employer of the job held today with the account the contact is filed under. Move the people who left, flag them, or keep them as champions who moved; never leave them on the old account as current.
- Clear the guesses. Remove values that were never confirmed, or mark them as unverified, so that nobody acts on them.
Leads follow the same logic in the other direction: a new lead has to be attached to its account before anything else happens. The page on lead to account matching shows how to do it on the company page instead of the email domain.
How do you deduplicate CRM records without merging the wrong ones?
CRM deduplication works on an identifier that names the company or the person, not on how they are written. Name matching fails in both directions: it merges different companies with similar names, and it misses the same company written three ways.
- Acme Incacme.comImported from a trade show list
- ACME Corporationacme.ioCreated by a rep
- Acmeacme.co.ukCreated by a form fill
Keep what made the duplicates look different, as attributes of the one record: the domains, the name variants, who created each copy. They are useful for matching what comes in later, as long as they are no longer the key.
For contacts, the profile URL plays the same role. Two contacts with the same profile URL are the same person, even with two different emails. People can edit their profile URL, though: when a stored URL comes back not found, resolve the person again from their work email before you decide anything.
Is CRM data cleansing the same as CRM data hygiene?
No. Cleansing is a project, hygiene is a routine. CRM data cleansing, or CRM data cleaning, repairs the records you already have, once. Data hygiene keeps them right afterward, by re-checking records as things happen to them.
A yearly cleanup fails for a simple reason: decay does not wait for it. Records start going out of date the day after the cleanup ends, and the people who just moved, the records most likely to be wrong, are exactly the ones a yearly pass checks too late. CRM data hygiene best practices come down to moving the check to the event, with a rolling schedule for the records no event touches.
How do you keep CRM data clean after the cleanup?
Tie every re-check to something that happens in the CRM. When a record is about to be used, check it then.
- A lead arrivesReverse Email Lookup/v2/enrichment/email-to-personWho it is, the current job, the account
- An account goes activeCompany Enrichment/v2/enrichment/companyIndustry, size, headquarters, domain
- A contact bounces or goes quietProfile Enrichment/v2/enrichment/personThe job held today, with its start date
- A deal opensWaterfall ICP/v2/search/waterfall-icp-keywordThe decision makers in their seats now
foundtrueWrite the value, with the date it was checked.
foundfalseMark it unconfirmed, with the date. Never present the stored value as checked.
- A lead arrives. Identify the person behind the email and attach the lead to its account before routing it. The page on inbound lead enrichment covers this step.
- Then the lead gets routed. Owner, territory and queue depend on the account you just matched. The page on lead routing covers the rules.
- An account goes active. Refresh the company profile before the first call: size, industry, headquarters.
- A contact bounces or stops answering. Check the job the person holds today before you look for a new address: if they moved, the old address belongs to a job they left.
- A deal opens. Look for the decision makers who are in their seats now, not the ones saved two years ago.
Between events, refresh a different slice of the database each week, so that every record gets checked over a few months, with recent job changers first. The CRM hygiene use case wires these triggers end to end.
What should you do when a source cannot confirm a record?
Write down that it could not, and stop presenting the stored value as checked. Not found means the source had no match. It does not prove the stored value wrong, so keep it, mark it unconfirmed with the date, and check it again later or from another key, such as the work email.
Three habits keep that information from getting lost:
- Store the date of the last check next to each field that changes: employer, title, email. A value without a date cannot be ranked against a fresher one.
- Keep three states, not two: confirmed, unconfirmed (checked, no match), never checked. Most CRMs only show filled or empty, which hides the difference between a value checked yesterday and one typed in years ago.
- Never fill a field to make it look complete. A guessed email bounces, a guessed title gets the wrong message, and both look like facts in every report built on them.
How do you measure CRM data quality?
Measure it on a sample, against a fresh source, by fault. A data quality score computed from how full the fields are measures completeness, not truth.
- Draw a random sample of a few hundred contacts and accounts, not the ones you worked on last week.
- Re-check each one against a fresh source: is the current employer still the account it is filed under, is the title still the one on file, does the account resolve to one company page.
- Count by fault: decayed, wrongly joined or duplicated, unconfirmed. Each has its own fix, so a single score hides what to do.
- Split the results by how recently each person started their job and by segment, and repeat the same sample design every quarter.
Three ratios are worth tracking between samples: accounts per company ID (above one means duplicates, unless you keep child accounts on purpose), the share of contacts whose current job was confirmed in the last 90 days, and the share of new leads matched to an account before routing.
Show the answer key
- Keep the contact on Globex: the email still works, so the record must still be fine.A working address is not a current job. A work mailbox usually stops working once its owner leaves, and until then every message reaches someone who no longer buys for Globex.
- Delete the contact: they no longer work at the account.You would lose someone who knows your product. A person who leaves a customer is often worth a call at the next company.
- Mark the Globex job as ended, attach the person to the Initech account, and look for the address that goes with the new job.The history stays true, the contact lands on the account they actually work for, and the old address stops being used as if it were current.
For builders
Everything above runs on a few calls: resolve the key, check the job held today, refresh the account. This part shows them, with a real lookup, and what to check before you ship.
How do you re-check a CRM record with the API?
Profile Enrichment (Person Enrichment in the API reference): a profile URL in, the person's whole career out, with the job they hold today marked as current.
Send
person_linkedin_url- The key you stored on the contact.
Read
found, person- The person, or found: false when no profile matches.
person.experiences[]- Each job with company_linkedin_id (a string, or null when the company has no page), job_start_date, job_end_date and job_is_current.
A company page URL in, the company profile out.
Send
company_linkedin_url- The page stored on the account. From a domain, resolve it first with POST /v2/enrichment/domain-to-linkedin.
Read
company.linkedin_id- The company ID, returned as a number here and as a string on a person’s jobs: the key to deduplicate accounts on.
company.industry, size, employees_on_linkedin, hq- The firmographics that routing and scoring read.
Each person or company found is one record, and a call that finds nothing counts nothing; plans and rate limits are on the BlitzAPI pricing page.
The check, in a few lines
The decision a re-check makes is small: is the job the person holds today at the account the contact is filed under?
// IDs come back as a string on a person's jobs and as a number on a company: compare them as strings.
const current = person.experiences.filter((job) => job.job_is_current && job.company_linkedin_id);
const stillThere = current.some((job) => String(job.company_linkedin_id) === String(contact.accountCompanyId));
if (!stillThere) {
// The person moved: end the old role, attach them to the account of their current job,
// and look for the address that goes with it.
}A real lookup
On November 10, 2026, we looked up someone in revenue operations at Datadog from their work email, masked here. Reverse Email Lookup returns the career in the same shape as Profile Enrichment. Two of the person’s jobs, both at Datadog, trimmed to the fields a re-check reads, with the dates masked to the year:
{
"found": true,
"person": {
"experiences": [
{
"company_name": "Datadog",
"company_linkedin_id": "1066442",
"job_start_date": "2026-••-••",
"job_end_date": null,
"job_is_current": true
},
{
"company_name": "Datadog",
"company_linkedin_id": "1066442",
"job_end_date": "2026-••-••",
"job_is_current": false
}
]
}
}Same company ID on both jobs: the contact is still on the right account. But the current job started in 2026, under a new title, so a CRM record saved before the change carries the old one.
Before you ship
- Backfill the company ID on every account and the profile URL on every contact before any refresh.
- Compare company_linkedin_id on the current jobs with the account the contact is filed under, as strings. A person can hold more than one current job.
- Store job_start_date and the date of the check: they tell you which records to re-check first.
- Record found: false as unconfirmed, with its date. Keep the stored value, but stop presenting it as checked.
- Merging accounts on their name or their email domain.
- Refresh a slice of the database each week rather than all of it at once, and keep each run within your monthly records.
Sources and method
- Written by
- The BlitzAPI team
- Published
- API calls run on
- B2B data provider benchmark, September 2026, BlitzAPI
- Benchmark methodology, BlitzAPI
- CRM hygiene playbook, BlitzAPI documentation
- Person Enrichment, BlitzAPI API reference
- Company Enrichment, BlitzAPI API reference
- Domain to LinkedIn URL, BlitzAPI API reference
The figures on decay come from BlitzAPI's September 2026 benchmark and are computed from the published report when the page is built: the answers of all the providers in that edition are added up, BlitzAPI's included, and no provider is shown on its own. The report names each provider and gives its own figures. BlitzAPI runs this benchmark and is one of the providers measured.
Acme, Globex and Initech, their records and the company ID shown for Acme are illustrative. The Datadog lookup in the builders section was run with the BlitzAPI API on November 10, 2026. It shows the company, the person's function and a title change in 2026; the name, the exact titles, the address and the dates are masked.