Data Quality Checks: How to Audit Your CRM and Contact Database
Data quality decides whether a sales email reaches the right person, whether a forecast matches what happens, and whether an AI assistant writes to the right job title. Bad contact data shows up as bounced emails, missed forecasts, and reps who call people who left the company.
Poorly managed contact data manifests itself through bounced emails, inaccurate forecasts, and representatives calling customers that have already left the company.
The seven data quality checks included within this document for a CRM or contact database will allow you to perform each check using a specific test; establish a target number for the result of the test; and identify possible fixes. The final outcome based on your testing will determine which option you need to select from: clean up the existing data, add missing data, or start over with new data.
What Is Data Quality in a Business Database?
Data quality is how well your records match the real world and how well they suit the job you give them. A contact with the right name and the wrong employer has an accuracy problem. A contact with a valid email and no job title has a completeness problem, because no rule can send that record to the right rep.
Also Read: How to Find Who Owns a Business? Real Ways to Search
What Do the 2026 Numbers Say About CRM Data?
Sales teams’ own estimate. Firmable, a company that sells prospect data, published a survey on September 9, 2026. With 222 B2B sales professionals, it’s estimated that 32% of the contact and account records in their CRM are inaccurate, incomplete, or outdated. In a companion Firmable survey of 220 sales professionals, 58% said they had found bad sales data only after contacting a prospect, and 14% said an AI agent could act on their current sales data without cleanup first. Both surveys measure what respondents believe, and no one inspected the records.
Revenue loss. Validity, a data quality software company, surveyed 602 CRM users in the United States, the United Kingdom, and Australia in 2025. It found that 76% say less than half of their organisation’s CRM data is accurate and complete, and 37% say their company loses revenue as a direct result of poor data quality.
Email decay. ZeroBounce, an email validation company, checked more than 11 billion email addresses in 2025 and found that at least 23% of an email list degrades within a year. The figure was 28% in 2024.
AI projects. Gartner’s July 2024 survey of 1,203 data management leaders found that 63% of organisations either lack the right data management practices for AI or are unsure they have them. Gartner predicts that through 2026 organizations will abandon 60% of AI projects that lack AI-ready data. An AI assistant that drafts outreach uses the fields in your CRM, so a wrong job title ends up in the email.
Inbox rules. Since February 1, 2024, Google’s sender guidelines ask anyone who sends 5,000 or more messages a day to Gmail accounts to keep the spam rate in Postmaster Tools under 0.10% and never reach 0.30%. Microsoft began enforcing similar rules on Outlook.com, Hotmail and Live on May 5, 2025, and Gmail stepped up enforcement in November 2025.
Also Read: Data Enrichment vs Data Cleansing: Which Does Your CRM Need?
The 7 Checks, Step by Step
The seven checks below put a number on your own file.
1. Are the Names, Jobs and Phone Numbers in Your CRM Correct?
Aim for at least 95% fully correct records. Test a random sample of 200.
→ Open each person’s company website or LinkedIn page.
→ Confirm the name, employer, and job title.
→ Call 20 of the phone numbers.
→ Count a record as wrong if any field is wrong.
For example, 14 wrong records out of 200 equal 93% accuracy. A result under 95% means the full file needs a pass to check emails, phone numbers and addresses.
2. When Was Each Record Last Checked?
No more than a quarter of your records should be older than 12 months. Every record needs a last-checked date for this test.
If your CRM has no such field, create one and fill it with the date of the last import or edit. Then filter for records past 12 months and begin with senior titles and your top accounts. Quarterly CRM cleaning keeps the old share small.
3. Which Fields Have Blanks?
Each field your team uses should be filled in at least 80% of records. Count the blanks field by field.
You can find missing email addresses, add missing phone numbers and fill in company size and industry as separate jobs.
4. How Many Duplicate Records Do You Have?
Keep duplicates under 3% of your records. Divide the duplicates you find by your total records.
Search in three passes:
→ Same email address
→ Same company website plus same last name
→ Same phone number, with spaces and country codes removed
Add a fourth pass for similar spellings so “Acme Inc” and “Acme Incorporated” land in one group. Duplicate records let two reps contact the same buyer and count one deal twice in the pipeline. When you merge two records, keep the one with the latest activity and copy over any fields the other holds. For small files, look for a merge tool inside your CRM. For bigger files, you can merge duplicates in bulk.
5. Do All Your Tools Write the Same Field the Same Way?
Each country, state, and job title should have one spelling in every tool. Compare one field across your CRM, email tool, and billing system.
A country field that holds “United States,” “USA,” “US,” and “U.S.” splits every regional report into four rows. A search for “VP Sales” misses “Vice President, Sales” and “VP of Sales,” although all three describe one buyer.
Three rules fix most of it. Use dropdown lists for country and industry. Write every phone number with its country code in one pattern, such as +1 212 555 0100. Keep a list that maps each title variant to one standard title. Names and addresses in other alphabets need extra care, and you can clean records written in other languages with a team that reads them.
6. Do Your Emails and Calls Reach People?
Keep hard bounces under 2% and spam complaints under the 0.10% Google limit above. A hard bounce is an email rejected because the address does not exist.
Check 1,000 email addresses with an email checking tool, then send a small test to the valid ones and record the results.
Amazon SES puts an account under review at a 5% bounce rate and may pause sending at 10%, so a 2% limit gives you room to fix problems first.
Dial 50 phone numbers and count how many reach the named person.
Remove hard bounces from every list right away. Before any large send, clean your email list.
7. How Many Contacts Fit the Customers You Want?
At least 30% of your records should match your target customer. Write down the target first: industry, company size, country, and job titles. Then count the records that match all four.
For example, 12,000 matches in a file of 80,000 contacts give a 15% fit. The other 68,000 records take up rep time, and on any platform that charges by contact count, they add to your bill.
Archive the poor fits and rebuild from lists cut to your target: contacts sorted by industry, contacts sorted by job title, contacts sorted by country and city and companies that use a specific tool.
Is Your Business Data Still Reliable? How to Read Your Results
Add up the checks you missed. These cut-offs are this guide’s rule of thumb, so adjust them to your list size and sales cycle:
→ 0 to 2 missed: the data is reliable enough to use. Remove duplicates, fix formats, and repeat the checks each quarter.
→ 3 to 4 missed, or accuracy between 85% and 95%: check the whole file, fill the blanks, and refresh the oldest records first.
→ 5 or more missed, or accuracy under 85%: replace the file. Cleaning cannot add contacts that match a customer your file never held.
When you replace or top up records, ask every provider these five questions and drop any provider with two weak answers. BizProspex publishes its answers on its own site:
Ask every provider you shortlist for a sample and run check 1 on it. BizProspex lists the data types you can sample, including funding, jobs, events, sanctions, and healthcare.
Also Read: How to Find Ready-to-Buy B2B Leads Without Cold Outreach
Frequently Asked Questions
How do I pick 200 random records?
Export your contacts to a spreadsheet, add a column with the formula =RAND(), sort by that column, and take the top 200. At 95% confidence, a sample of 200 gives a margin of about 4 points, near 93% accuracy, and up to 7 points at worst. A sample of 400 brings the worst case down to about 5 points.
What is a good data quality score for B2B contact data?
Set your own pass mark. The checks above use 95% accuracy, and BizProspex publishes a 98% guarantee. On a file of 50,000 records, each point of accuracy equals 500 contacts.
How do I measure how fast my data goes out of date?
Retest the same 200 records after six months and count the ones with at least one changed field. If 30 changed, your six-month rate is 15%. For email addresses, compare your rate with ZeroBounce’s 23% yearly figure.
What is the difference between cleaning, checking, and adding details?
Cleaning fixes what you have, such as duplicates, formats, and typos. Checking, also called verification, tests whether each value is true today. Adding details, also called enrichment, fills fields you do not have, such as phone numbers, company size, or the tools a company uses.
Does GDPR require accurate data?
Yes. Article 5(1)(d) says personal data must be accurate and, where necessary, kept up to date, and that inaccurate data must be erased or corrected without delay. The rule covers personal data of people in the EU.
Can AI clean my CRM on its own?
AI tools can flag duplicates and format differences. A contact’s current employer is a fact outside your CRM, so someone has to confirm it against a source.
Which check should I run first?
Start with check 1, since its result shows how much of the file needs work. Run check 6 before your next email send.
Who should own data quality in a company?
Name one person in sales ops or marketing ops to run the quarterly audit. That person schedules the checks, records the results, and reports the number of missed checks to the team.