What is data accuracy? Definition, causes and how to keep B2B data accurate

Data accuracy is the degree to which a data value matches the real-world fact it describes. In a B2B contact list, that means the job title says what the person does today, the email reaches their inbox, and the company is the one they actually work for. A record that still says "VP Sales at Acme" six months after that person left is complete, neatly formatted and wrong.

Most definitions you will find were written for data warehouses and finance tables. This guide applies the idea to the list a Sales Ops lead, a RevOps manager or a growth marketer uses to prospect: what counts as accurate field by field, what breaks it, how to check it on a sample you control, and how to keep it right once the list starts aging.

What is data accuracy? A definition for B2B contact data

Data accuracy answers one question: is this value true right now? Not "is the field filled" (that is completeness), not "is it in the right format" (that is validity), not "is it the same in every system" (that is integrity). A value is accurate when you could check it against the world and it would hold up.

Two things make data accuracy hard to manage in prospecting data. It is measured per value, not per file: a row can carry an accurate company name and an inaccurate title at the same time. And it decays. A phone number that was right when you collected it stops being right the day the person changes jobs, even though nobody touched the file.

Here is what accurate and inaccurate look like on the fields that matter for outreach:

None of the inaccurate values in that table would be caught by a format check. Every one of them looks valid. That is why data accuracy needs its own method, separate from cleaning and formatting.

Data accuracy vs data integrity: same goal, different work

The two terms get used as synonyms. They are not. Data integrity is about data staying whole, consistent and unaltered across its lifecycle and between systems. Data accuracy is about the value being true. You can have perfect data integrity on a file full of wrong values: every sync ran, no row got corrupted, and the titles are all eighteen months old.

The difference shows in how each one is protected. Data integrity relies on access controls, backups, checksums, validation rules between systems and governance over who can edit what. Accuracy relies on verification against an outside source, cleansing, profiling and refreshing values that change in the real world. A good integrity setup stops your CRM sync from overwriting a correct title with a wrong one. It does nothing about that title going stale on its own.

You need both. Integrity keeps correct values from being damaged once you have them. Accuracy work makes sure the values were right in the first place, and are still right today.

Data accuracy vs data quality (and completeness)

Data quality is the wider idea. It asks whether the data is fit for use, and accuracy is one of its dimensions, next to completeness (is the field filled), freshness (when was it last checked), consistency (is it written the same way everywhere) and uniqueness (does each record exist once). If you want the full model and how to run it as a process, the hub covers the dimensions of data quality.

The useful point in the data quality vs data accuracy debate is that a record can be accurate and still low quality. A verified email for a student, a retiree or a company far outside your ICP is accurate. It is also useless for your campaign. The reverse happens too: a list can look healthy on a dashboard, with 95% of fields filled, and hide a poor accuracy rate underneath.

Completeness is the dimension most often confused with data accuracy, because enrichment tools report fill rates. A filled field is not an accurate field. To see how the two numbers move against each other, and why pushing one tends to hurt the other, read coverage rate vs accuracy rate. Fill rate, match rate and cost per lead are enrichment KPIs rather than accuracy measures; they are covered in data enrichment KPIs.

Why accuracy matters when you prospect

Gartner estimates that poor data quality costs organizations an average of $12.9 million per year. On a sales team the cost shows up in smaller, more visible ways, and most of them trace back to data accuracy rather than to empty fields:

• Bounces. Each wrong email wastes a send, and enough of them damage the sender reputation of the domain every rep uses.
• Wrong person. A stale title sends your message to someone who no longer owns the problem, or no longer works there.
• Wrong segment. An outdated headcount or industry puts an account in the wrong sequence, with the wrong pitch.
• Wrong score. Lead scoring built on stale firmographics ranks the wrong accounts first, and reps spend their best hours on them.
• Lost trust. After a few bad calls, reps stop believing the CRM and keep their own spreadsheets, which makes the next problem worse.

Missing data is visible: you see the empty cell. Inaccurate data stays invisible until it fails in front of a prospect. That gap is why data accuracy deserves its own check instead of being folded into a general quality score.

What causes inaccurate data

Inaccurate data rarely comes from one big error. It piles up from eight ordinary sources:

1. Manual entry. Typos in names and domains, a phone number copied from a business card, a title lifted from an email signature that was already out of date.
2. Decay. People get promoted and change companies; companies rebrand, merge or move. The file stays the same while the world changes, which is why a refresh every 3 to 6 months on active accounts is the usual rule, and why enriching once a year leaves most of the year on stale values. The details per field are in how often to refresh B2B data.
3. Duplicates that diverge. Two records for the same person get updated separately. One has the new title, the other the old one, and your sequence may pick the wrong one.
4. Inconsistent formats. "US", "USA" and "United States" in the same column, or phone numbers without a country code. The value may be right but it reads wrong, and filters built on it return the wrong rows.
5. Silos and syncs. The same contact lives in the CRM, the marketing tool and a spreadsheet, and an update in one never reaches the others.
6. Migrations and system bugs. One mapping error during a CRM migration can shift values into the wrong column across thousands of rows.
7. Undated third-party values. A bought or enriched value with no date attached gives you no way to know whether it was true last week or three years ago.
8. Biased collection. A form that defaults to the first option in a dropdown, or a badge scan that records the person who registered instead of the person who came.

Examples of inaccurate data in a B2B list

Seven cases you will find in almost any prospecting file, and how each one gets caught:

• The departed CMO. The title still reads CMO, and the person left nine months ago. Caught by checking the public profile or the company's team page on a sample.
• The bouncing email. The address follows the right pattern, but the mailbox was closed. Caught by email verification before the send, or by the hard bounce after it.
• The catch-all trap. The domain accepts every address, so the email looks fine and may still never reach a human. Caught by a verification status that keeps catch-all apart from valid.
• The wrong domain. The company rebranded and the list still holds the old domain, or the parent group's. Caught by comparing the domain with the company's current website.
• The frozen headcount. A startup listed at 11-50 people has grown to 200. Caught by re-enriching the company and comparing ranges.
• The switchboard posing as a direct line. The phone field holds the reception number. Caught by a call on a sample, or by checking the number against the company's public contact page.
• The diverging twins. Two records for the same person with two different titles. Caught by a duplicate check that compares field values, not just names.

Every example has the same shape: the value is filled, plausible and wrong. None of them would fail a completeness report.

How to measure data accuracy on a sample

You cannot verify every value in a 20,000-row file. You can verify a sample properly, and a sample measured with care tells you more than any headline percentage a vendor puts on its homepage.

1. Draw a random sample of 100 to 200 rows from the segment you are about to work, not from the whole database. Accuracy varies a lot from one segment to another.
2. Pick a source of truth per field. The company's own website for domain and company name, the person's public profile or the team page for title and employer, an SMTP-level verification for email, a call for a direct phone.
3. Check each field on its own and mark it correct, incorrect or unverifiable. Keep unverifiable values out of the denominator instead of counting them as correct.
4. Compute the accuracy rate per field: accuracy rate = values verified as correct / values verified × 100.
5. Write down the date. A data accuracy measurement is a snapshot. The same sample checked six months later will score lower, even if nobody touched it.

Two corrections to the way this is often measured. The duplicate rate is not a measure of data accuracy: it measures uniqueness. Duplicates cause inaccuracy when their values diverge, which is why they sit in the causes above and not here. And the hard bounce rate tells you about the email field only. A list with 1% bounces can still be full of wrong titles.

The same sample is also the fairest way to compare two data sources: run both on the same 100 to 200 rows and compare accuracy field by field instead of trusting the fill rate. The full set of formulas, with benchmarks and the precision and recall view, is in the coverage and accuracy metrics guide.

Can data be 100% accurate?

Not for long, and not for contact data. You can reach 100% on a small sample on the day you check it. People keep moving, so the rate starts dropping the next morning. The realistic target is an accuracy rate you know, per field, with a date on it, and a refresh rhythm that keeps it above what your campaigns need.

Verify email accuracy before every send

Email is the one field you can test at scale without calling anyone, so it is where data accuracy work should start. A "found" email is not a deliverable one. Catch-all domains accept every incoming message without an SMTP rejection, which can inflate your fill rate while quietly hurting your deliverability.

For every batch, run a verification pass that sorts addresses into three groups:

• Valid: SMTP-confirmed, safe to send
• Catch-all: the domain accepts everything, moderate risk
• Invalid: guaranteed bounce, remove now

Targets after verification: hard bounce rate under 3%, catch-all rate under 15% in your active send list.

Derrick's Email Verification runs this check on a whole list in the web app, or from the Google Sheets sidebar if your list already lives in a spreadsheet. It costs 1 credit per email and is part of the paid plans. You get a three-tier list (valid, catch-all, invalid) and, on your sample, a real accuracy rate for the email field instead of a promise.

The mistake to avoid: confusing enriched data with verified data

Enrichment finds a value. Verification confirms that it is true. Treat them as two separate steps, both required. Skipping the second one is the most common way inaccurate data reaches a prospect: "found" emails that bounce at scale because nobody checked them, and a damaged sender reputation that takes weeks to repair.

How to improve data accuracy and keep it there

• Validate at entry. Required formats on forms and imports (email syntax, a fixed country list, phone with country code) stop part of the errors before they land.
• Verify before you use. Email verification before every campaign, and a sample check on titles before a sequence goes to a new segment.
• Deduplicate by comparing values. Merge twins and keep the most recent value. If your list is in Google Sheets, Derrick's Find Duplicates runs from the sidebar and flags repeated rows.
• Normalize the formats that change meaning. An inconsistent format makes a correct value read wrong: a country written three ways, a phone number without its country code. Fix those on the fields you filter and route on first. The limits to hold per field are in data hygiene thresholds per field.
• Re-enrich with a date. Refresh volatile fields (title, company, email) every 3 to 6 months on active accounts, and store the date of the last check next to each value. Enrich Leads costs 1 credit per profile and needs your LinkedIn account connected through the Derrick Chrome extension.
• Give each field an owner. Someone owns contact identity, someone owns firmographics. When everybody can edit a title, nobody is responsible for it being right.
• Monitor. Check accuracy on every import, and re-check a sample every month on the segments you are working. No campaign goes to a segment whose accuracy you have not measured recently.

If a large share of the file is already broken, start with a one-off B2B data cleansing pass, then put the habits above in place so the problem does not come back.

Where Derrick fits

Derrick handles the part of data accuracy that needs an outside lookup, from the surface that matches the job:

• Web app: import a CSV, verify the emails, re-enrich titles and companies. Nothing to install, and the default for working a list.
• Google Sheets sidebar: the same lookups on a list that already lives in a spreadsheet, plus Find Duplicates.
• MCP in Claude or ChatGPT: check one account or one contact in a conversation, right before a call.
• REST API: verify and refresh inside a CRM workflow or a scheduled job, from your own code or through Zapier, Make or n8n.

The free plan gives you 100 credits per month with no card, enough to re-enrich a test sample. Email Verification needs a paid plan, starting with Standard at €20 per month. API and MCP access come with the Plus plan at €47.50 per month.

Start with the free plan and measure data accuracy on your own sample.