Coverage vs Accuracy Rate: the Data Accuracy Metrics Behind B2B Enrichment

You enrich a list of 10,000 leads. Your provider's dashboard tells you: "92% coverage rate, 88% accuracy rate". Sounds great, right? Wrong question. The right question is: which of these two metrics matters more for YOUR use case, and what tradeoff are you willing to accept?

Coverage and accuracy are the two data accuracy metrics every enrichment dashboard reports. They look like cousins. They're actually opposing forces, and most B2B teams optimize the wrong one.

Coverage Rate vs Accuracy Rate: Precise Definitions

Before going further, let's nail down what each metric actually measures. Confusing them is the source of most bad decisions.

Coverage rate (also called fill rate)

Definition: percentage of rows for which the provider returned non-null data.

Formula: (rows with data returned / total rows submitted) × 100

Example: you submit 10,000 leads, the provider finds an email for 8,500 of them. Coverage = 85%.

Accuracy rate (also called precision rate)

Definition: percentage of returned values that are actually correct.

Formula: (correct values returned / total values returned) × 100

Example: among the 8,500 emails found, 7,650 are valid (the lead is reachable). Accuracy = 90%.

The trap: combining the two

Most providers communicate only their coverage rate. It's the metric that sells: "we cover 95% of LinkedIn profiles". But coverage without accuracy is worthless.

The real metric to track is effective rate: coverage × accuracy.

In our example: 85% × 90% = 76.5%. That's the percentage of leads where you actually have usable data.

Data Accuracy Metrics: 8 Formulas to Track on an Enriched B2B List

Coverage and accuracy are the headline numbers, but they are not the only data accuracy metrics worth computing. A data quality dimension is a property (accuracy, completeness, freshness); a metric is the number you compute to score it. On an enriched B2B list, eight formulas cover everything a dashboard needs.

Metric	What it answers	Formula	B2B example
Coverage rate	Did the provider return something?	rows with a value / rows submitted	8,500 emails on 10,000 leads = 85%
Accuracy rate	Is what came back correct?	correct values / values returned	7,650 valid on 8,500 = 90%
Effective rate	How many rows can I actually use?	coverage × accuracy	85% × 90% = 76.5%
Error rate	How much noise am I importing?	wrong values / values returned (1 minus accuracy)	850 wrong on 8,500 = 10%
Bounce rate	The field test of email accuracy	bounced emails / emails sent	above roughly 2%, deliverability starts to suffer
Validity rate	Is the value well formed?	values passing format rules / values returned	phone numbers in E.164, domains that resolve
Duplicate rate	Am I paying twice for the same contact?	duplicate records / total records	300 duplicates on 10,000 = 3%
Freshness rate	Was it true recently?	values verified in the last 90 days / values returned	a job title checked last month vs last year

A provider that exposes a confidence score per row lets you compute most of them without a manual sample. Read them in pairs. Validity is not accuracy: an address can be perfectly formatted and still belong to someone who left the company. And bounce rate is accuracy measured too late: it is the same number as the accuracy rate, discovered after it has already cost you reputation. For the step-by-step method to score these on your own base, see our guide on measuring data accuracy; for how fast values decay, read the page on data freshness.

Why this distinction matters in B2B prospecting

Each downstream play has a different tolerance to coverage vs accuracy. Knowing which one tolerates what changes the whole stack.

Cold email outreach: accuracy >> coverage

Sending an email to a wrong address has direct consequences:

• Bounce rate increases → your domain reputation falls
• You end up in spam → even your good emails don't land
• Inboxing rate collapses on the whole campaign

Rule: better to have 5,000 verified emails (99% accuracy) than 10,000 random emails (60% accuracy). The campaign on 5K will perform 3x better.

Account targeting (ABM): coverage >> accuracy

If you target 500 accounts on a multi-channel ABM campaign, you need to identify these accounts even if some signals are imperfect.

Better to know that "TechCorp uses HubSpot" with 75% confidence than to know nothing. You can verify later.

Rule: maximize coverage on the qualitative signals, accept the noise.

Lead scoring: accuracy >> coverage

If your scoring model is based on faulty data, it scores wrong. Result: you prioritize bad leads and ignore the good ones.

Rule: prefer to score 60% of your base with accurate data than 100% with mediocre data.

Market research / trend analysis: coverage >> accuracy

For statistical analyses (which technologies are gaining ground, which sectors are growing), you need volume. A 5% margin of error on a sample of 50K is acceptable.

Rule: volume above all, statistical accuracy is sufficient.

How to Measure Data Accuracy Metrics (Coverage and Accuracy) on Your List

Measuring coverage rate (easy)

Coverage rate is the simplest metric to measure. Most providers display it directly.

Manual method:

1. Count the rows submitted to enrichment
2. Count the rows that came back with non-null data
3. Calculate the ratio

Coverage rate breakdown: don't stop at the global rate. Segment by:

• Industry (SaaS often > 90%, manufacturing < 60%)
• Geography (US > 95%, Asia 50-70%)
• Company size (Enterprise > 90%, < 10 employees < 50%)
• Required data type (email vs phone vs intent data)

Measuring accuracy rate (harder)

Accuracy requires manual verification on a sample. There's no shortcut.

Standard methodology:

1. Take a random sample of 100-200 enriched rows
2. Manually verify each value:
   • For emails: send test, check bounce
   • For phones: call and verify identity
   • For job titles: check on LinkedIn
   • For tech stack: open the site and check the page source or the public DNS records
3. Calculate the rate: (correct values / total values verified) × 100

Frequency: re-test every 3 months minimum. Provider quality varies over time.

Data Accuracy Metrics vs Precision, Recall and F1 in Machine Learning

Search for data accuracy metrics and half the results talk about classification models. The overlap is real, and borrowing the vocabulary helps: an enrichment provider is a classifier that decides, row by row, whether to return a value.

• Precision = TP / (TP + FP). In enrichment terms, that is exactly the accuracy rate: correct values over values returned.
• Recall = TP / (TP + FN). If every row has a findable true value, recall is the effective rate: correct values over rows submitted.
• F1 = 2 × precision × recall / (precision + recall). One number that punishes a provider for being strong on one side and weak on the other.

Applied to the two providers from the mistakes section below: Provider A (78% coverage, 95% accuracy, effective rate 74.1%) scores an F1 of about 83%. Provider B (92% coverage, 70% accuracy, effective rate 64.4%) scores about 67%. The provider with the better-looking coverage loses by 16 points.

One trap: the ML metric called accuracy, (TP + TN) / total, is not the data accuracy rate. It credits the model for true negatives, which in enrichment means correctly returning nothing. That is a useful property (a provider that admits a miss is a provider you can trust, and one that returns a guess instead produces the false positives in enriched data that inflate coverage), but mixing the two definitions in a vendor comparison produces numbers that cannot be compared.

The convex tradeoff: why "more coverage" almost always means "less accuracy"

Providers face a structural choice:

• Strict mode: only return when confidence is high (90%+) → high accuracy, low coverage
• Permissive mode: return as soon as a signal is found → high coverage, lower accuracy

Most B2B providers choose permissive mode because it makes their dashboard look good. The 95% coverage rate is what sells.

The convex curve

The tradeoff isn't linear. Going from 80% to 90% accuracy might cost you 15% coverage. Going from 90% to 95% accuracy might cost you another 25% coverage. The marginal cost of accuracy goes up fast.

For most B2B use cases, the sweet spot is around 85-90% accuracy with whatever coverage that gives you. Don't chase 99%.

How to recognize a permissive provider

Signals that a provider is in permissive mode:

• Coverage rate > 90% across all geos and sizes (impossible without faking it)
• No confidence score per row
• No "last verified" date
• Marketing message exclusively on coverage, never on accuracy

Data Accuracy Metrics Benchmarks: Coverage Rate vs Accuracy Rate

A coverage or accuracy figure with no context is unreadable, and vendors know it. Three reference points make any published number interpretable.

The data type sets the ceiling. These are not one market with one difficulty level. A company domain is close to universally resolvable. A professional email sits materially lower. A direct mobile number is the hardest asset in B2B data by a wide margin, and any provider quoting the same figure for all three is quoting a blended average across an easy majority.

The market moves the number more than the tool does. Coverage on a US mid-market list and on a French SME list are not comparable measurements, whatever the vendor. Comparing two tools on two different lists tells you which list was easier, not which tool is better.

Freshness is a hidden term in both metrics. A record that was accurate at collection and has not been re-checked in a year is counted as covered and as accurate by most published methodologies, and it will still bounce. When a provider quotes accuracy, the question that matters is accurate as of when, and re-verified on what cycle.

Coverage and Accuracy by Data Type

The practical consequence of the ceiling effect is that you should never budget a workflow against a single blended rate. Break the expectation down by what you are actually asking for.

• Firmographics (domain, industry, headcount band, location): the easiest layer. Expect high coverage, and expect accuracy problems to come from staleness rather than from wrong matches. A headcount band from eighteen months ago is not wrong, it is old.
• Professional email: the middle layer, and the one where the coverage-accuracy trade-off is sharpest. A provider can lift coverage substantially by returning pattern-guessed addresses, and every point gained that way is a point lost on accuracy. This is precisely why finding and verifying are separate operations at Derrick, at 5 credits per email found and 1 credit per address verified.
• Direct phone: the hardest layer, and the one where per-result billing matters most. At 200 credits per number actually returned and nothing on a miss, a low-coverage segment costs you nothing to test, which turns an unknowable coverage rate into a cheap experiment rather than a purchasing decision.

Read that way, "what is your coverage rate" stops being a useful question and becomes three questions with three different answers.

The Question to Ask a Vendor Instead

Both metrics are self-reported, measured on a list the provider chose, under a definition the provider wrote. Rather than comparing two numbers produced under two undisclosed methodologies, ask three things that are hard to answer vaguely.

1. What am I charged for on a miss? Per-result billing makes the coverage rate your supplier's problem rather than yours. Per-lookup billing makes it yours entirely, and it makes an optimistic coverage claim expensive rather than merely inaccurate.
2. Is verification included or separate? If it is bundled, the accuracy figure and the coverage figure are being produced by the same pipeline and cannot be checked independently. Separate steps are auditable.
3. Can I run my own list before committing? A hundred rows from your actual target segment produces a number that means something for you. Every published benchmark was measured on a list that was not yours.

The third one settles most vendor comparisons in an afternoon, and it is the reason a renewing free plan is worth more than a marketing page. Derrick's free plan opens 100 credits per month with no card required, which is enough to measure both metrics on your own data before any decision.

Measuring Data Accuracy Metrics with Derrick: Web App, Sheets, MCP or API

The fastest audit is a sample of 100 to 200 rows from your real target segment. Derrick lets you run it wherever the list already lives, with the same per-result billing on every surface.

• Web app: import the CSV in the Derrick web app, run the email finder, then verification (both on paid plans, from STANDARD at 20 EUR a month). Rows with a value give you coverage; verified rows give you accuracy. Nothing to install.
• Google Sheets: if the list is already in a spreadsheet, the Derrick sidebar enriches it column by column and you compute the three rates with a COUNTIF next to the data.
• Claude or ChatGPT (MCP): for a spot check, ask your assistant to find and verify a dozen contacts through the Derrick MCP server and compare the answers with what your CRM holds.
• REST API: to re-measure every quarter without anyone remembering to, schedule the same sample through the API (available from the PLUS plan) and log the rates over time.

Because the Email Finder charges 5 credits per email actually found and nothing on a miss, a weak segment costs you nothing to test, which also caps the cost of bad data on the gaps: coverage becomes a measurement, not a purchasing gamble.

The 3 mistakes most teams make

Mistake 1: tracking only coverage

You optimize for coverage. Your provider proudly displays 95%. You're happy. Then your cold email campaign bounces at 18% and Google flags you as spam.

Fix: add accuracy to your dashboard. Even a rough estimate (sample of 50, manual check) is better than nothing.

Mistake 2: switching providers based on coverage alone

Provider A: 78% coverage. Provider B: 92% coverage. You switch to B. Your campaigns get worse.

Why? Provider A had 95% accuracy → effective rate 74%. Provider B has 70% accuracy → effective rate 64%.

Fix: always compare on effective rate (coverage × accuracy), not on coverage alone.

Mistake 3: setting the same accuracy threshold for all use cases

You set a 95% accuracy threshold globally. For cold email, that's right. For ABM target identification, you're losing too much data - 75% would do the job.

Fix: define accuracy thresholds per downstream use case, not globally.

Data Accuracy Metrics Thresholds: Coverage vs Accuracy Rate per Use Case

Here's the matrix to use when picking your accuracy/coverage threshold:

Use case	Min accuracy	Min coverage	Why
Cold email (outbound)	95%	40%+	Bounces kill domain rep
Cold call (outbound)	90%	50%+	Bad numbers waste SDR time
LinkedIn outreach	85%	60%+	Lower stakes, easier to verify in-app
ABM targeting	75%	80%+	You need account-level visibility
Lead scoring	90%	60%+	Bad data → wrong prioritization
Market research	80%	90%+	Stats need volume, tolerate noise
Intent / behavioral signals	60%	40%+	Inherently noisy, validate downstream

These thresholds are starting points. Adjust based on your downstream metrics - if bounce rate jumps, raise accuracy. If your TAM coverage feels too sparse, lower the threshold.

Key takeaways

• Coverage rate = how many rows enriched. Accuracy rate = how many of those are correct. They're not the same and they trade off.
• The metric that matters is effective rate: coverage × accuracy. A provider at 92% coverage and 70% accuracy ships less usable data than one at 78% coverage and 95% accuracy.
• Define accuracy thresholds per use case, not globally. Cold email needs 95% accuracy; ABM identification can live with 75%.
• Most providers optimize for coverage because it's the metric on the dashboard. Always sample-check accuracy yourself before signing.
• Re-measure accuracy every 3 months - provider quality drifts, especially after they raise funding.

Conclusion: the metric that matters is the one tied to your business outcome

Stop arguing about coverage vs accuracy in the abstract. Pick the downstream metric you care about (reply rate, meetings booked, pipeline coverage) and back-solve: which combination of coverage and accuracy maximizes that outcome?

For most outbound teams, the answer is: cap accuracy at 90%, then maximize coverage from there. For ABM teams, it's the inverse: maximize coverage, accept 75% accuracy, validate the top accounts manually.

Where to start:

1. Audit your current provider on both metrics (sample of 200 rows for accuracy)
2. Calculate effective rate (coverage × accuracy)
3. Match against the matrix above for your use cases
4. If gap > 10 points, change provider or strategy