Most of the cost numbers in circulation belong in the footnotes of a business case, not the headline. Two published numbers are worth quoting, each inside its limits. And the number that actually closes the case is one you produce yourself, from your own records, in about 30 minutes.
The most quoted answer comes from Gartner: poor data quality costs organizations at least $12.9 million a year on average, according to Gartner research from 2020. It’s a real number. It’s also six years old, and it covers data quality across a whole organization, not the CRM. That’s the pattern with most bad-data cost figures. They’re true in a narrow shape and get quoted in a wide one, and a CFO only needs one follow-up question to find the gap.
So here’s the position this paper takes. Most of the cost numbers in circulation belong in the footnotes of a business case, not the headline. Two published numbers are worth quoting, each inside its limits. And the number that actually closes the case is one you produce yourself, from your own records, in about 30 minutes. The test is at the end.
What does bad CRM data cost a company?
No published study has measured it for your company, and the surveys that come closest measure opinions, not audits. Validity’s State of CRM Data Management in 2025, a survey of 602 CRM users and stakeholders, found 37% reported losing revenue as a direct consequence of poor data quality. It doesn’t say how much. The same survey found 76% said less than half of their organization’s CRM data is accurate and complete.
Read those for what they are. The 76% is how CRM users rate their own data. The 37% is how many said bad data cost them revenue, with no amount attached. Both are good for one job: showing a CFO the problem isn’t unique to your team. Neither tells you what it costs you.
Gartner’s $12.9 million is the only dollar figure in the usual set, and it’s the oldest. It’s a cross-industry average for data quality in general, and it dates from 2020. If you use it, use it as context with “2020 research” attached, and don’t let it be the number your case rests on.
Quote two numbers and footnote the rest
If you’re building the case this quarter, two published numbers hold up under a CFO’s questions, because each measures something you can price or check yourself. One is a time cost. The other is a decay rate, used only for the slice it covers. The rest go in a footnote with their date and sample, or come out of the deck.
The time cost you can price with your own payroll
Salesforce’s State of Sales, 7th edition, a survey of 4,050 sales professionals across 22 countries published in February 2026, puts manual data entry at 13% of the average workweek.
That’s the one number on the list you can turn into dollars without borrowing anyone else’s math. Take 13% of a rep’s week, multiply by your rep headcount and your loaded hourly cost, and you have a weekly figure built from a published rate and your own numbers.
Know its limit before you put it on a slide. It measures where reps’ time goes, not whether what they type is right. The two problems are linked, and why good reps enter bad data covers how the time cost turns into the quality cost.
The footnotes
The other figures you’ll see are fine to cite in the shape below and risky in the shape they usually travel in.
| Number | Quote it as | Don’t quote it as |
|---|---|---|
| At least $12.9 million a year on average (Gartner) | 2020 research on the cost of poor data quality to organizations | What bad CRM data costs a company today |
| 76% say less than half of their CRM data is accurate and complete (Validity, 2025) | How 602 CRM users and stakeholders rate their own data | An audit result |
| 37% report losing revenue to poor data quality (Validity, 2025) | A share of respondents who say they lost revenue | A revenue loss figure |
| 55% report conflicting pipeline signals; 42% lack formal governance frameworks for data consistency and accountability (Clari Labs, January 2026) | A Censuswide survey of 400 North American enterprise leaders | A finding about mid-market or startup companies |
If your company is smaller than enterprise, say so about the Clari Labs numbers before someone in the meeting says it for you.
How fast does CRM data go stale?
Two recent measurements put a number on decay, and each covers a narrow slice. ZeroBounce’s Email List Decay Report for 2026, based on the email data it verified from January to December 2025 (more than 11 billion addresses processed), found at least 23% of an email list decays within one year, and only 62% of emails submitted in 2025 were valid.
Lusha measured 140,284 US sales leaders at VP and C-suite level on September 12, 2026 and found 12.25% changed roles over twelve months and 25.67% over twenty-four. Lusha set that against the roughly 30% a year contact decay statistic it says data vendors cite. Its own measured number for that group is well under half of it.
Both companies sell data or verification, so treat both as vendor research. And they measure different things. ZeroBounce measures email lists, not whole CRM records: a contact can keep a valid inbox and still sit at the wrong company, owned by a rep who left. Lusha measures role changes among senior US sales leaders, which says little about your mid-level champions or your accounts outside the US.
Inside those limits, the pair is useful for two specific arguments. A sequence built on last year’s export is sending a real share of its volume to addresses that no longer work. And the exec contacts on your biggest open deals deserve a check before the forecast call, because a fair share of senior sales leaders move within two years.
Which numbers should you leave out?
Drop any figure that has no named source, date and population. Drop any figure stretched past what its source measured, like an email decay rate presented as CRM decay, or a 2020 cross-industry cost presented as this year’s CRM cost. A CFO who traces one number back to nothing will discount the rest of the deck with it.
Headlines are a softer case. HBR’s 2017 headline, “Only 3% of Companies’ Data Meets Basic Quality Standards”, from Tadhg Nagle, Thomas C. Redman and David Sammon, is fair to quote as the article’s headline with its date attached. As a current CRM benchmark, it’s stretched past what you can defend.
A short list of numbers in their narrow shape carries more weight than a long list in their recycled shape. If you’re torn on one, leave it out.
The cost isn’t in the bad records
Here’s the part most business cases miss. A lot of the bad data in a CRM costs close to nothing. A wrong fax number on a dormant contact doesn’t hurt anyone. Cost shows up where a workflow reads a field and acts on it: routing reads territory or industry, the forecast reads close date and amount, and sequences read the email address. A duplicate contact is clutter until two reps work the same buyer in the same week, which is the pattern behind HubSpot duplicate contacts.
That changes what you should measure. “What share of our data is dirty?” is the question the surveys answer, and it’s the wrong one for a business case. The better question is what share of the fields your workflows read are wrong. A CRM can be messy overall and fairly cheap, or mostly clean and expensive, if the dirty part happens to be the close date on open deals.
It also tells you where to start cleaning. Not the whole database. The handful of fields something downstream depends on. If you’re putting AI agents or scoring on top of the CRM, those are the same fields they’ll lean on, and what AI-ready data means lays out the bar they have to clear. The forecast side of the same problem is in the dirty reality of CRM data.
How do you measure the cost of bad data in your own CRM?
Borrowed numbers start the conversation. Your own sample closes it. HBR’s 2017 article suggests a simple exercise: assemble 100 data records completed by your department and work through each. Here’s a 30-minute version aimed at the fields that carry cost, so you walk out with your own number instead of someone else’s.
- Pick four fields a workflow reads (5 minutes). One that drives routing (owner, territory or industry), close date and amount on open deals, and the primary contact’s email. Those cover the routing, forecast and outreach costs from the section above.
- Pull 50 records that are in play (5 minutes). Open deals and the contacts on them, exported to a spreadsheet. Skip the dormant part of the database. It makes the number look worse and matters less.
- Mark each field (15 minutes). Right, wrong, missing or can’t tell. Some are wrong on sight: a close date already in the past on an open deal, a blank industry, an email your activity log shows bounced. Mark “can’t tell” and move on rather than researching.
- Count and trace (5 minutes). Count the records with at least one wrong or missing field, and next to each, write what the workflow did with it: routed to the wrong rep, sitting in the forecast on a date that can’t be true, a sequence step that bounced.
That gives you two things a published number can’t: a count from your own pipeline, and a list of what it broke.
For a reference point from our own work: across the 150+ B2B companies we’ve worked inside, roughly 7 in 10 had material CRM data accuracy problems. Our methodology section explains how we define and count that. It’s context for your test, not a substitute for it.
One thing to do this week: block 30 minutes before your next pipeline or forecast meeting, run the test on your open deals, and bring the count into the meeting before anyone builds a slide.
You have two ways into the CFO’s office. Take a borrowed number and spend the meeting defending its source, its date and its sample. Or spend half an hour on your own records and walk in with a number that came out of your own pipeline.
If you’d rather not count by hand, AeolusGTM’s free CRM Diagnostic runs 40+ automated checks across data quality, pipeline hygiene, freshness, reachability and ICP fit. It’s read-only and works with HubSpot, Salesforce and Pipedrive.