Every quarterly review has the same moment. Marketing presents one pipeline number. Sales presents another. Finance quietly has a third, in a spreadsheet nobody else has seen. Everyone pulled their number from “the system”, and everyone is right by their own definition.
Here is the hard truth: the CRM is not lying on purpose. It is simply recording what people put in, and what people put in is incomplete, out of date and occasionally made up.
How bad it is, in the companies’ own words
In 2025, Validity surveyed 602 CRM users and administrators in the US, the UK and Australia. 76% said less than half of their CRM data is accurate and complete. 37% said they lose revenue directly because of poor data, and for a quarter of companies that loss is at least 10% of annual revenue.
What CRM users say about their own data
Share of 602 CRM users and administrators in the US, the UK and Australia, 2025
The detail that should worry a board most: 37% said staff fabricate data to keep decision-makers happy. And 46% have nobody working full time on CRM data quality; a third do not know who is responsible at all.
The 2026 edition of the same survey asked about AI. Only 21% of marketers said their CRM data is very well prepared to support it. Yet 78% of C-suite respondents said they had acted on an AI recommendation they later suspected was wrong because of bad data.
That is the new risk. Bad data used to produce a bad report that someone might question. Now it feeds a model that acts on it, at speed, with confidence.
Unity Software learned what that costs. On its first-quarter call in May 2022, chief executive John Riccitiello said the company had lost the value of part of its training data, partly from ingesting bad data from a large customer, and that a fault had reduced the accuracy of its Audience Pinpointer targeting tool. He put the impact at about $110 million in 2022. The fix was monitoring, alerting and rebuilding the data before retraining the models.
Most companies will never see their own cost written down that clearly. It is there all the same, spread across missed forecasts, wasted campaigns and deals worked twice.
The teardown: five checks, live
In this episode we take a CRM apart on screen and run five checks. Each one takes minutes, and you can run them on your own system this week.
- Duplicates. How many companies and contacts appear more than once? Duplicates split activity across records, so nobody sees the full history and two reps chase the same buyer.
- The forecast fields. Pick the five fields your forecast depends on, such as stage, amount, close date, owner and next step. What share of open deals has all five filled in, and changed in the last 30 days?
- Stale stages. How many deals sit in the same stage with no activity for 60 days or more? Those are not pipeline. They are hope.
- Records with no owner. Every account, contact and deal needs one named person. A record nobody owns is a record nobody fixes.
- Where it came from. Can you trace a closed deal back to the campaigns and conversations that created it? If not, every budget decision is a guess.
What fixing it looks like
Cleaning the database once does not work. It decays again within months, because people change jobs and teams keep typing. What works is three habits.
- Rules at the door. Required fields, picklists instead of free text, and validation when a record is created, not a clean-up after.
- A named owner. One person accountable for data quality, with the authority to change how the system is used.
- A weekly look. The five checks above, on one screen, reviewed by sales, marketing and RevOps together.
Trust in the numbers is not a software feature. It is an operating habit, and it starts with someone owning it.
How far does your data leak? The Revenue Engine Health Check scores your revenue data and RevOps alongside four other areas, and shows which one to fix first. Five minutes, free.
See also our one-minute video, One funnel, four versions of the truth.