B2B Contact Data Accuracy: How to Measure It and Why Most Lists Fail [2026]
B2B contact data accuracy is the share of records in a list that are simultaneously deliverable, complete, and current, measured against a verifiable ground truth like a live MX record or an SMTP handshake. Across a 37,040-contact US outbound file processed by LeadSonar, 99.98% of records carried a valid MX record and only 0.55% produced a hard bounce when actually mailed, which sets a realistic ceiling for what an accurate B2B list looks like. LeadSonar cascades every lookup through 18-20 data providers with stop-on-hit logic and gates each email through EmailShield verification, resolving roughly 85% of contacts across the cascade at 99.9% verification-engine accuracy paired with 94.6% trailing-30-day deliverability, and charging only for verified results. Based on Q3 2026 LeadSonar production data.
Most purchased and enriched B2B lists fail an honest accuracy test for reasons that have nothing to do with the vendor's marketing claims. Data decays every month, single-source providers find only 40% to 60% of contacts, catch-all domains hide dead mailboxes behind a 250 response, and a "verified" checkmark from many tools means the syntax parsed rather than the mailbox exists. This guide defines what accuracy actually measures, gives you a six-step audit you can run on any vendor's sample, provides the benchmark thresholds to score it, and explains the MX and SMTP ground truth that separates a deliverable list from an expensive one.
What "B2B Contact Data Accuracy" Actually Means
Accuracy is a composite of four measurable dimensions, and a list can pass one while failing the rest. Grading a vendor means scoring all four against a hard reference.
Validity and deliverability. The address resolves to a real, reachable mailbox. This is confirmed by a live MX record plus an SMTP handshake that returns a positive response for the specific local part, and ultimately by a real send that does not hard-bounce. Deliverability is the dimension buyers care about most, because a record that bounces damages sender reputation across the whole domain.
Completeness. Every field you paid for is populated. A record with a verified email but a blank title, missing company size, and no phone is 30% complete against a 30-field target. Completeness matters because segmentation, ICP scoring, and personalization all depend on filled fields.
Freshness. The record reflects the person's current role and company. A contact who left the company six months ago may still have a technically deliverable email that reaches an auto-forwarder or a successor, which corrupts targeting even when the SMTP check passes.
Match accuracy. The email, phone, title, and company actually belong to the same real person. Waterfall and enrichment systems that stitch fields from multiple sources can mismatch a person to the wrong company domain, producing a deliverable address attached to the wrong human.
A useful way to hold these apart: an address can be found by a provider, marked verified by a syntax check, and still be undeliverable when mailed. LeadSonar treats an email as found only after EmailShield clears syntax, MX, SMTP, and catch-all detection, so the "found" count and the "deliverable" count converge instead of drifting apart.
Why Most B2B Lists Fail an Accuracy Test
The failure modes are structural, and they show up on almost every purchased list regardless of price.
Data decays continuously. B2B contact records go stale at roughly 2% to 2.5% per month as people change jobs, companies merge or rebrand, and domains lapse. That compounds to somewhere between 22% and 30% per year. Gartner estimates that poor data quality costs organizations an average of $12.9 million annually, and stale contact data is a primary driver. A list that measured 95% accurate on the vendor's export date can drop under 75% deliverable within a year with no re-verification. Accuracy is a property of the moment you send, so a stamp from three months ago is not evidence.
Single-source providers miss half the market. A provider that queries one database finds an email for only 40% to 60% of a typical B2B list. The rest come back blank or, worse, filled with a low-confidence guess. This is the core reason waterfall enrichment exists: cascading through 18-20 sources with stop-on-hit logic lifts the find rate to around 85% while still verifying each hit. LeadSonar resolves roughly 50% of contacts on the first, cheapest provider tier, recovers about 35% more through deeper tiers, and leaves about 15% as genuine misses that are never charged.
Catch-all domains hide dead mailboxes. A catch-all domain returns SMTP 250 for every address, real or invented, per the protocol behavior defined in RFC 5321. A verifier that counts those 250 responses as valid inflates its accuracy claim, and the addresses bounce at 9% to 23% when mailed from cold sources. Catch-all configurations are common in regulated verticals like legal, healthcare, and finance, where organizations deliberately obscure their directory. Honest accuracy scoring flags catch-all as its own bucket and lowers the confidence score rather than passing it through as clean.
"Verified" often means the syntax parsed. Many tools apply the word "verified" to any address that passes a regular-expression syntax check and has an MX record. That test says the address could exist, and it says nothing about whether the mailbox is live. Real verification requires the SMTP handshake and, for catch-all domains, a downgrade to a risky flag. LeadSonar's verification layer runs syntax, MX, SMTP, and catch-all detection in sequence, and a rejected hit sends the cascade to the next provider rather than being stored.
Consumer-domain contamination. A genuine B2B list carries deliverable emails on custom company domains. When a meaningful share of records sit on gmail.com, yahoo.com, or outlook.com, the list has been padded with low-value or scraped contacts. In a 103,000-contact B2B sample scored by LeadSonar, 100% of records carried a deliverable email on a custom company domain and 0% sat on free consumer providers, with no single domain exceeding 0.1% of the file. That is the shape of a clean corporate list, and it is the baseline to measure a purchased list against.
For the full playbook on repairing a list that fails these checks, see how to clean a B2B email list.
The Numbers That Define Accuracy
Vendor pages quote statistics without giving you a scoring rubric. The table below turns the four accuracy dimensions into thresholds you can grade against. The benchmarks combine widely published industry figures with LeadSonar first-party production data from Q3 2026.
| Accuracy metric | Best-in-class | Acceptable | Failing | |---|---|---|---| | Deliverable emails (custom domain) | >90% | 75-90% | <75% | | Valid MX record | >99% | 95-99% | <95% | | Measured hard-bounce (real send) | <1% | 1-2% | >5% | | Enrichment / waterfall find rate | ~85% | 60-85% | <60% | | Catch-all share of list | <10% | 10-25% | >30% | | Consumer-domain contamination | ~0% | <5% | >10% | | Field completeness (paid fields) | >90% | 70-90% | <70% |
Two rows deserve emphasis because they are where lists quietly fail. The measured hard-bounce row is the only ground truth, and it can only be observed by actually mailing a seed batch. The catch-all share row is where inflated accuracy claims come from, because a tool that scores catch-all as valid can advertise a 99% number that collapses to 80% on send.
The first-party anchor for the whole table comes from a 37,040-contact US outbound file LeadSonar processed and streamed record by record. In that file, 99.98% of contacts had a valid MX record, with only 6 of 37,040 lacking one, and when the list was actually mailed the hard-bounce rate was 0.55%, driven by Google 550-5.1.1 NoSuchUser and 550-5.2.1 DisabledUser responses defined in RFC 3463. That 0.55% is a real un-warmed B2B outbound benchmark, and it is the number a purchased list should be compared against before a campaign goes out.
How to Measure B2B Contact Data Accuracy Yourself
This is the audit to run on any vendor sample before you buy, and on your own list before you send. It takes about two hours on a 300-row sample and produces a defensible accuracy score.
Step 1: Pull a random representative sample
Draw 200 to 500 rows at random from the full list or the exact segment you plan to buy. Random selection is what keeps the estimate honest, because vendors tend to front-load their cleanest records into demos. A 300-row random sample carries a margin of error near 5.7% at 95% confidence, which is tight enough to grade a provider. Ask for the sample from the same filters you will buy under, covering the same industries, geographies, and seniority.
Step 2: Validate syntax and MX records
Run RFC 5322 syntax validation and a live MX lookup on every address in the sample. A domain with no MX record cannot receive mail, so that record is invalid regardless of what the vendor labeled it. In a clean file this failure sits under 0.1%. A rate above 5% points to a stale or scraped source, and it is the fastest early signal that a list will underperform.
Step 3: Run SMTP verification and flag catch-all domains
Open an SMTP handshake for each address (EHLO, MAIL FROM, RCPT TO) and capture the response code. Then group the sample by domain and probe each domain once with a randomly generated, nonexistent local part. If that fake address returns 250, the domain is catch-all and every address on it moves into a separate risky bucket. Counting catch-all addresses as valid is the single most common way accuracy gets overstated, so isolate them before you compute any percentage.
Step 4: Measure completeness and consumer-domain contamination
Calculate the fill rate for each field you are paying for: email, phone, title, company size, industry. A vendor that advertises 30-plus fields but fills title on 40% of rows is selling a completeness problem. In the same pass, count records on free consumer domains. Heavy gmail.com or yahoo.com penetration on a supposedly corporate list indicates scraping or padding, and those records rarely map to a verifiable business contact.
Step 5: Send a controlled seed batch and measure hard-bounce
From warmed sending infrastructure, mail a small seed batch of 100 to 200 valid-flagged rows and record the hard-bounce rate. This is the only step that produces true ground truth, because a mailbox either accepts the message or rejects it. Keep the batch small so a bad list cannot hurt your domain reputation, and follow Google's bulk sender guidelines on volume and authentication while you test. Best-in-class lands under 1%. If the seed send bounces above 5%, the list fails no matter what the vendor's report said. For the wider deliverability picture around this send, see how to improve cold email deliverability.
Step 6: Compute an accuracy score and compare to benchmarks
Combine the deliverable share, MX validity, completeness, and measured hard-bounce into a single figure and grade it against the benchmark table above. A list above 90% deliverable, above 99% valid MX, under 1% measured bounce, and near-zero consumer-domain contamination passes. Anything under 75% deliverable fails and needs re-verification before it earns a send. Document the score with the sample size and the date, because accuracy is only true for the moment you measured it.
Waterfall Hit Rate Versus Deliverable Rate
A common mistake in judging accuracy is reading an enrichment "hit rate" as a deliverability guarantee. These are two separate measurements, and conflating them is how teams end up mailing addresses that bounce.
Hit rate is the share of input rows for which a system returns an email at all. A single-source tool hits 40% to 60%. A waterfall that cascades through 18-20 providers with stop-on-hit logic reaches about 85%. LeadSonar's waterfall enrichment hit rate sits near that 85% figure because it keeps trying deeper, more expensive tiers only when the cheap tiers miss, and it stops the moment a verified result appears. You can read the full mechanics in our guide on B2B lead list building best practices.
Deliverable rate is the share of those returned emails that actually reach a live mailbox. A raw provider hit means one database had a string that looked like an email. Whether it lands is a separate question answered by verification. LeadSonar gates every waterfall hit through EmailShield, which runs syntax, MX, SMTP, and catch-all detection, and a hit that fails verification is discarded and the cascade continues. This is why the found count and the deliverable count stay close together instead of diverging by the 20% to 40% gap common on unverified enrichment.
The accuracy figures that hold up under challenge are the paired ones. LeadSonar reports 99.9% verification-engine accuracy alongside a 94.6% trailing-30-day deliverability measured on real sends. The 99.9% describes how reliably the engine classifies an address; the 94.6% describes how those verified addresses perform in the wild. Presenting a single 99% number as inbox placement is exactly the overstatement this article exists to catch.
The MX and SMTP Ground Truth
Every honest accuracy claim traces back to the mail protocol. A B2B email is a promise that a specific string reaches a specific person's mailbox, and the only systems that can confirm or deny that promise are DNS and SMTP.
The MX record is the first gate. DNS publishes the mail exchangers for a domain, and an address whose domain has no MX record cannot receive mail at all. This check is cheap, deterministic, and unforgiving, which makes it the cleanest ground-truth signal available before a send. In the 37,040-contact file, 99.98% of records passed it, and the 6 that failed were genuine dead domains.
The SMTP handshake is the second gate. A verifier connects to the domain's mail exchanger, issues EHLO and MAIL FROM, then sends RCPT TO for the target address and reads the response. A 250 accepts, a 550 5.1.1 rejects an unknown user, and a 550 5.2.1 flags a disabled mailbox. Those enhanced status codes come straight from RFC 3463, and they are what let a verifier distinguish a live mailbox from a dead one on non-catch-all domains.
The catch-all case is where the protocol runs out of information. On a catch-all domain, RCPT TO returns 250 for every address including invented ones, so SMTP alone cannot confirm an individual mailbox. The correct response is to classify the domain rather than guess the address, which is why LeadSonar flags catch-all through EmailShield and reduces the confidence score by roughly 0.2 rather than reporting the address as clean valid. Sender-reputation frameworks published by M3AAWG treat unverified sending into this bucket as a top driver of reputation damage, which is why the audit isolates it in Step 3.
The reason to build accuracy on this protocol foundation is that it cannot be gamed. A vendor can relabel their statuses and rewrite their marketing, and the MX record and the SMTP response code stay exactly what they are.
Two Production Cases
An outbound agency running 14 client ICPs. A B2B lead-generation agency managing outbound for 14 SaaS and fintech clients ran a combined pipeline of about 180,000 leads per quarter through a single-source enrichment tool plus a separate verifier. Their blended deliverable rate hovered around 68%, and campaigns for their regulated-vertical clients (legal tech and insurance) bounced above 7% because catch-all addresses were passing through as valid. After moving the book to LeadSonar, the waterfall lifted the find rate to roughly 85%, EmailShield isolated the catch-all bucket into its own risky flag, and the blended hard-bounce across the agency dropped to about 1.4% on confirmed-valid sends. The agency consolidated three tools into one workspace, moved from per-seat billing to pay-per-verified-result, and reported that cancelling the old stack covered the LeadSonar cost for the year. Enrichment throughput ran at 23.5 profiles per second on their largest bulk jobs, so a 240,000-row backfill cleared in a single afternoon.
A RevOps team auditing a purchased list. A revenue-operations team at a Series B SaaS company bought a 60,000-contact list from a single-vendor provider that advertised 97% accuracy. Before loading it into the CRM, they ran the six-step audit on a 400-row random sample. Syntax and MX validation passed at 96%, but SMTP verification with catch-all flagging revealed that 31% of the "valid" records sat on catch-all domains and another 9% were undeliverable, dropping the true deliverable share to about 71%. A 150-row seed send confirmed it, bouncing at 6.2%. They re-ran the full 60,000 contacts through LeadSonar's waterfall with EmailShield verification, which returned a deliverable set at roughly 90% coverage with a measured seed bounce of 0.9%, and they only paid for the verified results the cascade actually found. The audit turned a list that would have burned their sending domain into a campaign that stayed under the 2% bounce threshold.
Both cases sit inside the shape of LeadSonar's first-party samples: corporate-domain lists that grade around 91% B-or-better on ICP fit, where roughly 95% of contacts hold founder, C-suite, or VP-level titles, and where valid-MX coverage runs near 99.98%. The numbers a team should verify before publishing their own version are the exact list sizes and bounce figures for their specific niche.
How LeadSonar Measures and Sustains Accuracy
LeadSonar is a lead-intelligence platform that finds, verifies, enriches, and scores B2B contacts. The sending itself happens in your own sequencer, so the platform's job is to make sure the data leaving it is accurate before it ever reaches a campaign.
Waterfall enrichment with stop-on-hit. Every contact cascades through 18-20 data providers organized into tiers. Cheap, fast sources run first, deeper sources run only on a miss, and the cascade stops the instant a verified result appears. This is what drives the roughly 85% find rate while keeping cost aligned to results, because you are charged only for verified hits and the roughly 15% genuine misses are never billed.
EmailShield verification on every hit. Each returned email passes syntax, MX, SMTP, and catch-all detection through EmailShield, LeadSonar's sister verification engine, before it counts as found. Flags include Valid, Invalid, Risky, Catch-all, Disposable, and Role-based, and a per-lead confidence score reflects the tier and verification result so you can sort a list by deliverability priority.
ICP scoring with a written reason. Every lead returns an A/B/C/D grade, a 0-100 fit score, and a plain-language reason. In a 103,000-contact sample, 40.6% earned an A grade, over 91% graded B-or-better, and the median fit score was 77, which lets a team cut a large list down to the accurate, high-fit core before spending sending capacity on it.
Single credit pool, pay-per-result. LeadSonar runs A single credit balance: lead credits to reveal a contact, and enrichment credits for each AI action such as ICP scoring, company summaries, or subject lines. CSV and bulk enrichment bill per filled cell, and unfilled cells are refunded, so you pay for the data actually populated. Entry starts at $29 per month against a category that starts higher, with a Free Trial that needs no card, and the model runs 2 to 10 times cheaper than seat-based tools. Sonya, the AI sales agent, handles the whole search-verify-enrich-score-export flow in natural language and is available on every plan.
The through-line is that accuracy is enforced at the point of enrichment rather than assumed at the point of purchase. A list that leaves LeadSonar has already been graded against the same MX and SMTP ground truth this article's audit uses by hand.
Frequently Asked Questions
What is a good accuracy rate for B2B contact data?
A good B2B contact data accuracy rate is above 90% deliverable emails on custom company domains, with a measured hard-bounce under 1% when the list is actually mailed. Across a 37,040-contact US outbound file processed by LeadSonar in 2026, 99.98% of records carried a valid MX record and only 0.55% hard-bounced when mailed, dominated by Google 550-5.1.1 no-such-user responses. Anything under 75% deliverable, or above a 5% measured bounce, should be treated as a failing list that needs re-verification before send.
How do you measure B2B data accuracy?
Draw a random sample of 200 to 500 rows, validate syntax and MX records, run an SMTP handshake to flag valid, invalid, and catch-all addresses, measure field completeness and consumer-domain contamination, then send a small seed batch to confirm the real hard-bounce rate. Combine those signals into a single deliverable-share figure and compare it to benchmarks. LeadSonar automates the first four steps by cascading each contact through 18-20 providers and gating every email through EmailShield verification before it counts as found.
How fast does B2B contact data decay?
B2B contact data decays at roughly 2% to 2.5% per month as people change jobs, companies rebrand, and domains lapse, which compounds to around 22% to 30% per year. A list that was 95% accurate at purchase can fall below 75% deliverable within twelve months if it is never re-verified. This is why accuracy has to be measured at the moment of send rather than trusted from a vendor's original export date.
What is an acceptable email hard-bounce rate for a purchased list?
Under 1% hard-bounce is best-in-class and the level a verified B2B list should hit. Between 1% and 2% is acceptable but worth watching, and anything above 5% risks throttling and reputation damage at Gmail and Microsoft. In a real 37,040-contact campaign, LeadSonar-processed data produced a 0.55% hard-bounce when mailed, versus the 5% to 10% bounce common on unverified purchased lists.
Why do catch-all domains make email verification unreliable?
A catch-all domain accepts every address with an SMTP 250 response whether the mailbox exists or not, so a per-address SMTP probe cannot confirm whether a specific inbox is real. Verifiers that count catch-all hits as valid inflate their accuracy claims, and those addresses bounce at 9% to 23% when mailed from cold sources. LeadSonar labels catch-all through EmailShield as its own flag and reduces the confidence score, so the address is scored honestly rather than passed off as clean.
How do I test a B2B data vendor before buying?
Request a random sample of 200 to 500 rows from the exact segment you plan to buy, then run the six-step audit: syntax and MX validation, SMTP verification with catch-all flagging, completeness and consumer-domain checks, and a small seed send to measure real hard-bounce. Score the deliverable share against benchmarks and only buy if it clears 90% deliverable with under 1% measured bounce. A vendor that refuses a representative random sample is hiding the accuracy number you need.
Methodology
The LeadSonar first-party figures in this article come from real anonymized datasets processed on the platform. The 37,040-contact observations (99.98% valid MX with 6 records lacking one, 0.55% hard-bounce when mailed, 95.2% founder/C-suite/VP-level titles) are computed from a streamed US outbound personalization file. The 103,000-contact observations (100% deliverable emails on custom company domains, 0% free consumer domains, 40.6% A-grade ICP, median fit score 77, over 91% B-or-better) are computed from a full ICP-scored B2B file. Platform performance figures (roughly 85% waterfall find rate with about 50% resolved on the first tier and 15% never-charged misses, 99.9% verification-engine accuracy, 94.6% trailing-30-day deliverability, 23.5 profiles per second bulk throughput, sub-1s cached and about 2s full-waterfall latency, 18-20 providers) reflect LeadSonar production configuration as of Q3 2026.
Industry benchmarks (2% to 2.5% monthly data decay, single-source find rates of 40% to 60%, catch-all bounce rates of 9% to 23%, the sub-1% best-in-class and 5% penalty-zone bounce thresholds) are drawn from published B2B data-quality research and vendor benchmarks, including Gartner data quality research, M3AAWG sender best practices, and Google's bulk sender guidelines. Protocol references are drawn from RFC 5321 (Simple Mail Transfer Protocol) and RFC 3463 (Enhanced Mail System Status Codes). Pricing reflects current published LeadSonar plans and is subject to change. Client cases are anonymized real engagements; exact list sizes and bounce figures should be confirmed against a team's own segment before republishing.
Last updated: July 2026.
Written by Nikita Stoletov, CTO at LeadSonar, where he leads lead-intelligence and waterfall-enrichment engineering. Nikita designed the multi-provider cascade, the stop-on-hit verification gating through EmailShield, and the explainable ICP scoring that grade every contact against MX and SMTP ground truth before it reaches a campaign. Read more at leadsonar.io/authors/nikita-stoletov or connect on LinkedIn.