What Does One Contact Cost? 15c to 83c at the Same Tier

Here's the trap. You put four data tools side by side at 15, 25, 47 and 83 cents a contact, buy the 15-cent one, and load 2,000 records. Nine hundred come back empty. Of the 1,100 that return something, a chunk bounce on the first send, and your domain reputation takes the hit rather than the vendor's.

The advertised price is a numerator. It buys a row. What your reps need is a row that is accurate and deliverable, and the share of rows that clear that bar is the denominator nobody sells you. Divide one by the other and the 15-cent contact from LeadMagic can cost more than the 25-cent one from Prospeo.

It is like buying eggs by the dozen from stalls that do not let you open the box. The price on the sign is per dozen. What you cook with is the number that are not broken, and the stalls differ on that more than they differ on price.

This page does not sell or resell any tool on it. Every price below is attributed to the vendor's own page with the date it was read. Nothing here is a measurement. This site has not run a record sample against any vendor, and every cost figure is computed from published list rates. Where a question can only be answered by running the product, this page says so and tells you what to measure.

So here is what the page contains: the nominal cost of one contact at every vendor whose rate can be read, the division that turns it into a real unit cost, the crossover point at which a dearer vendor becomes the cheaper one, what the verification layer adds, what all twelve published accuracy claims actually claim, and the 200-record procedure that gets you your own denominator for about $52.

What does one usable contact actually cost?

Between 15 and 83 cents before you divide, at entry pricing, from the six vendors in the data category that publish enough to compute it. That is the honest headline and it is also the least useful number on this page, because it prices a row rather than a usable row.

After you divide, the range runs from roughly 19 cents to over four dollars for the same six vendors, depending entirely on what share of your list each one can actually serve. The division is larger than the price difference. It is larger than the price difference by about a factor of five, which is why the ranking every roundup publishes is a ranking of the numerator.

Nobody can give you the denominator, including this page. It is a property of your list crossed with their database, not a property of the vendor, and the one thing every published figure in this category has in common is that it was measured on somebody else's list.

What this page can do is give you the numerator to three decimal places, show you exactly how much denominator it would take to overturn each pairwise comparison, and hand you a procedure that produces your own denominator in a morning.

Credit, coverage, accuracy: three words doing one job badly

Now that the fraction is on the table, the four terms it rests on are worth one sentence each, because the category uses all four loosely and two of them get quoted interchangeably.

A credit is not a unit. It is whatever the vendor says it is, and what it buys varies by a factor of eleven across the same six pricing pages. At UpLead one credit reveals a contact with the work email and the mobile direct dial together. At BetterContact the mobile costs ten credits and the email costs another one.

Coverage, also sold as match rate, is how much of your list comes back with anything at all. Accuracy is how often what came back is right. They move independently, a database can be excellent at one and poor at the other, and vendors quote whichever is higher without always saying which one it is. Both are in your denominator and they multiply.

A catch-all, also labelled accept-all, is a domain configured to accept mail to any address, so a verifier cannot tell whether the specific mailbox exists. It is neither a yes nor a no. On plenty of B2B lists it is a large fraction of the results, and how each vendor classifies it silently moves both your bill and their published accuracy figure.

That last one is the load-bearing definition on this page. A vendor that counts an undecidable catch-all as valid will report a higher accuracy figure than one that does not, on byte-identical data, without either of them saying anything false.

Why the fill rate, not the truth, is what gets scored

The reason the denominator is unpublished is not that it is hard to measure. It is that measuring it is against the interest of the party holding the data, and someone who worked inside that business has described the scoring mechanism in public.

On a 2022 Hacker News thread with 807 points (item?id=32399949, comment 32405067, 2022-08-09), a commenter who says they left the data industry in 2020 and whose NDA expired in 2021 writes: "Data quality is typically assessed at the 'Does this data field have a value for this line item' level. That means data vendors are financially incentivized to make shit up about you as much as they can get away with."

That is one person's characterisation, not a finding of fact, and they name no employer. It is worth quoting because it describes a mechanism rather than an outcome, and the mechanism is checkable against the pricing pages: every vendor on this page meters and bills on rows returned, and not one of them meters on rows that turned out to be right.

The same commenter adds the part that explains the decay problem later on this page: "the data goes stale faster than the accuracy of the data becomes relevant." Again, one person on a public thread. But a buyer whose security or procurement team goes looking will find that thread, and it is better to have read it first.

The commercial consequence follows directly from the meter. If you are billed per row returned and the vendor is scored on fill rate, then a row containing a plausible guess earns revenue and a blank row does not. Nothing in that arrangement is fraudulent. It simply means the number you are buying and the number you care about are not the same number.

The one vendor that publishes its own refresh ceiling

One vendor does put a number on the staleness problem, and it is the most useful sentence any data vendor publishes about its own denominator. LeadIQ's pricing-page FAQ states, read 2026-08-25: "You query a live dataset of 750 million+ professional profiles, refreshed every month. Up to 200 million records are re-verified as people change companies or roles."

Take the arithmetic at face value and 200 million of 750 million is 26.7 percent, and the words "up to" make that a ceiling rather than a rate. So at best around a quarter of the database is re-verified in a given month and at least 73.3 percent is not. That is vendor-reported and should be treated as a ceiling, not a floor, which is precisely why it is worth having: the ceiling is already below what the category's accuracy claims imply.

LeadIQ deserves credit for two other things on the same page. It labels coverage and accuracy as different quantities rather than blending them, quoting "85 % global match-rate (70 % in EMEA and climbing)" for mobile coverage and "Accuracy > 90 %" for verified numbers. And it defines what verified means, as numbers that "pass our AI-driven carrier checks". It still does not say how the 90 percent was arrived at, so the method is missing like everyone else's, but the quantities are at least named separately.

The arithmetic, written out

One formula, and everything else on this page is an input to it or a consequence of it.

The credits-per-contact term is the mobile multiplier, and it is published, so it belongs in the numerator where you can read it. The coverage and accuracy terms multiply rather than average, which is why the division bites harder than people expect. Sixty percent coverage at ninety percent accuracy is not seventy-five percent. It is fifty-four.

The one distinction that changes the formula

There are two billing shapes in this category and they take different denominators. Charge-per-attempt vendors spend a credit when you ask, whether or not anything comes back, so coverage sits in the denominator and misses inflate your unit cost directly.

Charge-on-success vendors spend a credit only when a result returns and passes verification. FullEnrich states that credits are spent only when data is found and passes verification, that a lookup returning nothing costs nothing, and that a landline found instead of a mobile costs nothing either. BetterContact states the same and adds that a credit is charged only after successful verification. Both are vendor-reported and both are stated in writing on the vendors' own pages.

For those two, coverage drops out of the cost denominator entirely. It does not drop out of your problem, because a contact you cannot buy at any price is still a contact you do not have, but it stops inflating the per-unit figure. That is the actual product being sold under the word waterfall, and it is worth more than the mechanism the marketing describes.

What one contact costs before you divide

The unit is one person with a verified work email and a mobile number, priced at each vendor's cheapest paid tier, at list, assuming the lookup succeeds. Prices read from each vendor's own pricing page on the dates shown.

A spread of 5.5 times across the six that publish a comparable figure, and 73 times if LeadIQ's page is read at the setting it loads on. LeadIQ shows Pro at $200 a month against a credit slider whose rungs run 200, 500, 1,000 and upward to 6,750, and it opens on 200. At that rung a credit is $1.00 and a contact with both an email and a phone is 11 credits, so $11.00, read 2026-08-25.

That is the slider trap this site has now found on four vendors. Dropcontact opens on EUR 79 against a EUR 29 floor, FullEnrich on $55 against $29, and Hunter and LeadMagic both default to annual billing. The default view is not the floor, and a comparison table built by loading pricing pages and reading the first number is measuring the vendor's preferred anchor. Whether $1.00 a credit holds at the 6,750 rung cannot be read without running the page's JavaScript, so this page publishes only the rung it loads on and says which one that is.

The mobile multiplier moves the answer further than the headline price does, and no vendor publishes the comparison because it reads badly in exactly one direction. UpLead has the highest per-credit rate on the table at 58.2 cents and the lowest multiplier at one, so it comes last of nine tools on cost per email and fourth on cost per complete contact. Nothing about UpLead changed between those two sentences. Only the question did.

What happens to the ranking when you divide

Here is the same nominal column carried through four denominators. The denominators are illustrative bands, modelled from stated assumptions and not measured, because this site has not run a sample against any of these vendors and nobody neutral publishes one.

Read across a row and the division costs more than any negotiation will win you. LeadMagic's contact moves from 18.5 cents to 75 cents without its price changing at all, a factor of four, on a spread of denominators that is entirely ordinary for B2B lists. No discount in this category is worth four times.

Read down a column, though, and the order never moves. That is the honest limit of the table above: a denominator applied equally to everyone rescales without reordering. The reordering comes from the denominators being different per vendor, which is the next section and the only part of this that is genuinely first-party work.

How much better does the dearer vendor have to be?

This is the number to take away. For any pair, the dearer vendor becomes the cheaper vendor once its usable-row rate beats the cheaper vendor's by the ratio of their prices. That ratio is computable from published rates alone, with no measurement required, and it tells you which parts of the ranking are fragile.

So the finding is narrower and more useful than "the ranking inverts". The gap between the two cheapest vendors is erasable by the denominator, and the gap between the cheapest and the dearest is not. A 1.63x swing in usable rate is an ordinary difference between two databases on one segment. A 5.5x swing is not, because no vendor runs at 15 percent usable while another runs at 83 on the same list.

Which means the decision splits cleanly in two. Choosing between LeadMagic at 15 cents and Prospeo at 24.5 cents on price is a mistake, because a single morning of testing can settle it the other way. Choosing between LeadMagic and BetterContact on price is not a mistake, as long as you are comparing like with like.

The last clause is doing real work. The two vendors at the bottom of that table bill on success, which takes coverage out of their denominator, so the comparison is not like with like and the price ratio overstates their disadvantage. That is the case for a waterfall and it gets its own arithmetic further down.

The verification layer is part of the denominator's cost

You cannot compute accuracy without paying somebody to check, and the checking has its own meter with its own hidden unit. Per-address rates at three volumes, read first-hand.

At a thousand addresses, four of these charge the same eight tenths of a cent, which tells you the entry rung is a market convention rather than a cost. Past that they diverge hard. NeverBounce's Growth subscription reaches $0.00159 at a hundred thousand, and MillionVerifier's slider shows 50,000 for $89, which is $0.00178 each, and a million for $449, which is $0.000449 each. That last figure is roughly a ninth of what the same vendors charge at a thousand, so volume moves verification pricing by nearly an order of magnitude more than it moves contact pricing.

Hunter is on this table because people compare it here, and it should not be. It is a finder that verifies as a convenience, charging half a credit a check, which works out at 1.2 cents. That is two and a half to nearly eight times the dedicated verifiers. Use it to find addresses, not to clean a list.

Does a catch-all cost you a credit?

This is the question that decides both halves of the fraction at once, and the vendors split four ways on it. It moves your bill, because catch-alls can be a large minority of a B2B list. It also moves every accuracy claim in this market, because a catch-all counted as valid is a free point of accuracy.

Clearout is the only one that publishes the whole rule. Its pricing guide states that an unknown status "will not be charged" because it is not a definitive outcome, while valid, invalid and catch-all results are "considered as billable service action and charged accordingly". Its consumption table repeats it. That is vendor-reported and it is checkable on the page.

MillionVerifier states the most buyer-favourable rule of the seven: "you only pay for good and bad emails and not for risky (unknown and catch-all) emails", read 2026-08-25. It also claims to be the only verifier that exempts catch-alls, which is a vendor claim about competitors' products and is not evidence about them. What can be said independently is narrower and still useful: Clearout's own page confirms Clearout bills catch-alls, so at least one competitor does.

NeverBounce publishes no equivalent rule at all. Every relevant help article was read and none exempts accept-all or unknown results from billing; the only stated exemption is deduplication. Its definition that one credit equals one verification implies all five result codes bill, but the company does not print that, so we are not reporting it as policy. The absence is the finding.

What NeverBounce does state is the part that costs twice. It guarantees no more than three percent of your emails will bounce, then excludes from that guarantee any address categorised as invalid, accept-all, unknown or disposable. So a catch-all almost certainly consumes a credit and definitely sits outside the accuracy promise you are paying for.

What every vendor claims, and what any of them define

Twelve published claims, set against the only question that makes a percentage comparable: did the vendor say how it was measured?

Eleven vendors publish a percentage and not one publishes a protocol. Two get partway. MillionVerifier says its 99 percent is "verified against billions of real-world sends", which names a basis without naming a method, and adds that "independent tests rank us #1" without naming the tests. Dropcontact names a sample size of 20,000 contacts, and is also a vendor benchmarking itself and declaring itself the most effective solution on the market. Both are vendor-reported and both should be read as ceilings rather than floors.

Two entries are worth more than the other ten. RB2B publishes 15 to 20 percent on basic resolution, which is not a marketing figure and is the only number on the table that cost the vendor something to disclose. Vector publishes two different numbers, roughly 10 percent in its help centre against 25 to 35 percent on its chief executive's blog, which is a useful reminder that these figures are produced by marketing departments rather than by instruments.

The most-cited apparently independent test is not independent either. It put nine databases against the same 500-record sample and produced a range of 65 to 98 percent, and it was run by Cleanlist, which sells data. That does not make the result wrong, and a 65 to 98 spread is more useful than any single vendor's number. It does mean the one neutral-looking data point in this category is not neutral, and there is nothing behind it.

What ZoomInfo tells the SEC that it does not tell buyers

One vendor in this market is public, which means once a year its lawyers describe the data business to a regulator in language that has to survive a securities claim. ZoomInfo's Form 10-K for FY2025, filed 2026-02-12 under CIK 0001794515, is the most candid document in the category and almost nobody in it cites the thing.

On the accuracy guarantee, the filing states that its "standard contract with customers includes a quality guarantee pursuant to which a customer would have the right to terminate its subscription and we could be obligated to reimburse certain payments if the accuracy of our data were to fall below a certain threshold". That is a genuine, enforceable accuracy commitment, and it is stronger than anything else on this page.

The threshold is not disclosed. Not in the 10-K, not on the pricing page, which carries no price either. So the one vendor in this market that contractually guarantees accuracy is also the one that will not tell you what it is guaranteeing, and you cannot compute a risk-adjusted price from a commitment whose trigger is a blank. Ask for the number in the redline, because it exists.

The contributory network, and what a free tier is for

The phrase "contributory network" appears in exactly six 10-K filings in SEC full-text search, and all six are ZoomInfo's (checked 2026-08-25). It is the company's own term for the mechanism by which its users supply the data it sells, and the filing describes it plainly.

"We provide a 'freemium' product, ZoomInfo Lite, which provides users with a set amount of monthly credits to use to access our contact data and, if the user elects, they may also participate in our contributory network in order to receive additional monthly credits." The filing then lists as a risk factor that Lite users and paying customers "may cease to participate", citing "their sensitivity to sharing information within our contributory network".

Read that next to the free-tier arithmetic further down this page. The credits are the consideration, the contact data in your address book is the price, and the company has told its investors that the arrangement depends on users continuing to find it acceptable. That is a fair trade if you know you are making it, and the pricing page is not where you find out.

The filing also names the regulatory direction of travel: California established the Delete Request and Opt-out Platform following the Delete Act, SB 362, "which launched on January 1, 2026, allowing consumers to submit a single deletion request to all registered data brokers", with New Hampshire and others expected to follow. Deletion at scale is a coverage problem, and coverage is in your denominator.

What the courts and regulators have been doing to data brokers

This is the part that shows up in month six rather than month three, and it bears on the denominator because every deletion, suppression or geographic carve-out removes rows you already paid for.

Apollo.io's corporate entity is Zenleads, Inc., which is how its litigation is docketed and why searching its brand name finds nothing. Four suits, from CourtListener RECAP, checked 2026-08-25:

Every row above is an allegation. A filed complaint is not a finding, none of these has been described here as one, and only a ruling would upgrade any of them. What the pattern shows is not that any vendor did anything, but that four separate plaintiffs in two districts across two and a half years thought a contact database was worth suing over, and that the Atlas Data Privacy campaign has aimed the same theory at the data supply chain underneath several tools on this page.

The two Illinois cases from 2019 matter because they are the older shape of the argument: that publishing a person's profile to sell a subscription uses their identity commercially. Whatever a reader thinks of that theory, its practical consequence for a buyer is suppression, and suppression comes out of coverage.

The European side, and why it is thinner than it should be

European data protection authorities have taken action against B2B contact brokers over legitimate-interest sourcing, and those decisions carry dates and amounts. This page does not cite a specific one, and the reason is worth stating rather than papering over: the search budget for this pass was exhausted before any decision could be read at the authority's own site, and repeating a fine from a law-firm summary would be exactly the chain-of-custody failure this site refuses elsewhere.

What can be said from a primary document read first-hand is ZoomInfo's own risk language, which names GDPR, the UK's post-Brexit transposition, India's Digital Personal Data Protection Act of August 2023, the CCPA as amended by the CPRA, and "similar comprehensive privacy laws adopted in a growing number of other states (approximately twenty as of January 2026)". A vendor telling investors it faces twenty state regimes is a vendor expecting suppression volume.

Why the denominator differs per provider: the waterfall

A waterfall asks one data provider for a contact, and if nothing comes back it asks the next, down a list of providers until something is found or the list runs out. Commercially the distinguishing feature is the charge-on-success rule, not the sequence.

That structure is the direct answer to why two vendors quote the same nominal price and return different amounts of usable data. A single-source tool has one database's coverage. A waterfall has the union of several, minus whatever its verification step strips out on the way through.

It costs about three times as much per email for that reason. At entry pricing read in August 2026, a work email costs about 2.5 cents at Hunter or Prospeo, 5.8 cents at FullEnrich and 7.5 cents at BetterContact, a range of 2.4 to 3.1 times. That multiple is where most people stop reading, and it is the wrong number, because you are not choosing between 800 contacts at 5.8 cents and 800 at 2.5 cents. You are choosing between 500 contacts and 800.

The 80 percent is what FullEnrich claims for itself and is a vendor figure rather than an independent test. On that assumption FullEnrich costs $34.15 more and returns 300 more contacts, so the extra contacts cost 11.4 cents each. Both waterfalls are several times the single-source rate per email, and both are cheap for a contact you could not otherwise have reached at all.

When does a waterfall stop being worth it?

When your existing tool is already good. The better it is, the fewer contacts the waterfall adds, and the same premium spreads over fewer of them. This is the inversion that the per-email comparison hides completely.

At 75 percent the arithmetic has turned over completely. An extra contact costs 56 cents, which is more than buying a whole contact with a verified email and a mobile number outright from the cheapest vendor in the category at 15 cents. At that point you are paying a premium to fill in the last few rows of a spreadsheet.

So the honest answer to whether you need a waterfall is another question, and it is the same question this whole page keeps arriving at: what is your current hit rate on your real list? Not on a demo list, not on the vendor's sample. Most buyers have never measured it, and it is the only input that decides this.

Three catches the waterfall pages do not lead with

• Verification is not total. FullEnrich states that 80 percent of catch-all domains are deep-checked, which means 20 percent are not, and whether those bill as a find is not answered anywhere on its site.
• The cheapest tier may not include catch-all verification at all. BetterContact's $15 Starter carries a dash against that row in its own comparison table, while an FAQ higher up the same page says catch-all validation incurs no extra charge. The FAQ is describing a feature Starter does not have.
• Rollover is capped. BetterContact rolls credits over up to your subscribed amount, so a 5,000-credit package banks at most 10,000. FullEnrich rolls over three months on monthly plans and twelve on annual.

Clay gets asked about here and belongs in a different row. It is an orchestration surface rather than a metered waterfall: it runs across providers you may be paying for separately, so the cost depends on what sits underneath. It starts at $185 a month against $15 and $29 for the two dedicated waterfalls, and this site's own note on it is blunt. If nobody on the team will spend a fortnight learning it, the credits expire unused.

Measure your own denominator: the 200-record protocol

Everything above is arithmetic over published rates. This is the part that produces a number nobody can publish for you. It costs about $52 at prices read in August 2026 and takes roughly half a day.

Two hundred records is enough to tell a 60 percent hit rate from a 90 percent one clearly enough to act on, and small enough to run on a single $15 plan. No confidence interval is offered here, because this site has not done that statistics and inventing one would be the same unearned precision the rest of the page is arguing against.

The seven steps

• 1. Build the sample from the awkward part of your ICP. Small companies, non-US domains, non-technical job titles, the segments you already suspect are thin. Keep company name, person name and domain, and nothing else, so you are testing the vendor rather than your own existing data. A vendor's demo list is selected to succeed and your best-known accounts flatter every tool equally.
• 2. Decide first whether you are testing email or phone. They are different tests at different prices. 200 credits buys 200 emails or 20 phone numbers at a 10x multiplier, and 20 results is not a sample. Testing phone coverage across the same 200 records needs around 2,000 credits, which moves you from a $15 plan to a $49 one.
• 3. Run it through two vendors, not one. A single result tells you nothing about whether 62 percent is good. One single-source tool and one waterfall is the most informative pairing, because the gap between them is the thing you are actually deciding whether to buy.
• 4. Count three numbers separately. How many rows came back with anything (coverage). How many came back with a mobile. How many survive verification and reach a human (accuracy). The category reports all three as one number and they are not one number. Coverage times accuracy is your denominator.
• 5. Verify through a third party, never the tool that supplied the data. Clearout's 100 free credits cover half the sample and NeverBounce at $8 a thousand covers the rest. Record the catch-all count separately, because that single number decides which verifier is cheaper for your list.
• 6. Send to it and count hard bounces. This is the only step that produces accuracy rather than a claim about accuracy, and it is the step everyone skips.
• 7. Compute overlap between your two vendors. If more than 70 percent of the rows one returns are also returned by the other, you are looking at substantially one dataset sold twice, and a second subscription buys you very little.

The pass/fail thresholds

• Coverage below 60 percent on your ICP means the vendor's headline contact count describes somebody else's ICP.
• Email accuracy below 85 percent puts you over most deliverability tolerances at volume.
• Direct-dial rate typically lands 10 to 30 percent, and must be counted as mobile numbers over rows returned, not as "a phone number is present", because switchboard numbers inflate that to near 100 percent.
• A fresh sample bouncing at five or six percent is behaving normally, not badly. Judge against decay, not against the vendor's claim.
• If the waterfall finds thirty points more than the single source, that gap is worth about eleven cents a contact and is a straightforward buy. Five points more and you are paying a premium to complete a list you already have.

What this procedure cannot detect. Recency. A correct-looking record for somebody who changed jobs four months ago passes every check above, verifies clean, and bounces the moment their old mailbox is finally deleted. It also cannot detect behaviour at your volume: 200 records is a smoke test, and a match rate measured on 200 will not hold at 50,000.

And if both vendors come back low, the problem may be your ICP definition rather than the vendors. Two independent databases missing the same segment usually means the segment is genuinely under-covered, which is worth knowing before you build a quarter's pipeline plan on it.

What the free tiers will and will not let you test

You can run most of the protocol above for nothing, but only if you pick the right free tiers, and the constraint that matters is not on any plan card.

Three of fifteen have a free allowance that renews and lets you export: Prospeo at 100 credits a month, Hunter at 50, and Persana at 50. Everything else is a one-time grant, a trial, or a plan you cannot get data out of. A one-time grant is a trial with the clock removed, which is genuinely useful for testing and useless for ongoing work, and four vendors present it under the same heading as a recurring allowance.

The export restriction is the test that removes half the recurring tiers, and it is the thing least likely to be on the plan card. Snov.io states that export is available on "all plans except Trial", and the renewable free plan is the Trial. Store Leads makes export a Pro feature, so its cheapest exporting plan is $250 a month. In both cases the restriction lives in the comparison table further down the page rather than on the tier itself.

LeadMagic is the most honest entry and has no free tier at all. It says in plain words that it no longer offers a free-credits programme and that new accounts choose a paid plan at signup, offering instead a full refund of the first subscription payment within fourteen days, one per organisation, explicitly excluding annual plans. For a buyer wanting to test properly that beats three of the free tiers above, because it comes with the whole product and unrestricted export.

What breaks in month three

The denominator you measured in week one is not the denominator you have in month three, and four separate mechanisms move it.

Decay, which is the big one

B2B contact data goes stale as people change jobs, at a rate usually quoted between 2.1 and 3 percent a month. That range circulates widely across vendor blogs, the figures conflict, and no traceable primary study sits behind any version of it, so treat it as a working assumption rather than a measurement. Run it forward on a list sold as 95 percent accurate, at 2.5 percent a month:

So a list sold as 95 percent accurate is roughly a 70 percent list twelve months later, and the vendor's claim was true when they made it. That single piece of arithmetic reconciles most of the gap between what these companies advertise and what buyers experience, without requiring anybody to be dishonest.

A practitioner put a blunter number on the same effect. On a 2019 Ask HN thread (item?id=21617138, 13 points, 11 comments), a commenter writes: "in b2b (target audience professionals in mid market companies) the 'lifespan' of an email address is no more than 2 years. People change jobs all the time." That is one person's characterisation, not a finding of fact, and a two-year half-life implies a monthly decay near the top of the published range rather than the bottom.

The three quieter ones

• Credits expire on a schedule nobody reads. Clearout and Bouncer state theirs never expire. NeverBounce's pay-as-you-go credits expire twelve months after purchase. HubSpot credits expire monthly. Prospeo serves two contradictory answers to the rollover question at random from its own bundle, so read your account terms rather than its pricing page.
• Bounces cost more than the credits did. A commenter running bulk email on a 2019 thread (item?id=20577416, comment 20581802) puts it at "a rule of thumb that every 'spam' click costs me (as a bulk email sender) about $10". That is one sender's stated figure from 2019 and not a market rate, but the direction is right: at 15 cents a contact, one complaint costs what 66 contacts did.
• Suppression removes rows you paid for. California's DROP platform launched 2026-01-01 and lets a consumer delete from all registered data brokers with one request, per ZoomInfo's own 10-K. Deletions are permanent and they come out of coverage without a refund.

The prices you cannot read at all

A price is only a price if the vendor shows it to a buyer, in a currency, for a stated period. Six vendors on most shortlists fail that test, and naming them is more useful than filling the gap.

Apollo and Prospeo have both previously burned this site by shipping React payload markers that a naive read takes for prices. A string like $1 or $L15 in a page's serialised state is a framework reference, not a currency amount, and it is the single most common way a wrong price enters a comparison table. Every figure on this page was read from rendered text.

This matters more than it sounds. A price you find quoted for ZoomInfo or Cognism almost certainly traces to a directory or to a competitor's comparison page, which is the one place a gated vendor's number reliably appears and the one place it must never be taken from.

Who should stop reading here

If you send fewer than about 500 emails a month, none of this arithmetic will pay for the morning it takes. The absolute sums are small, the difference between the best and worst vendor on 500 contacts is around $340 a year, and your time is worth more than that. Buy the cheapest tier with a free allowance, use it, and come back when volume makes the fraction matter.

If your reps only email and never dial, ignore the mobile multiplier entirely and buy on cost per email, where Hunter and Prospeo sit at roughly 2.5 cents and both have free tiers you can test on today. Half of this page is about a column you do not need.

And if you are shopping for one of the gated vendors because somebody senior named it, the arithmetic here cannot help you, because the numerator does not exist publicly. What can help is the ZoomInfo accuracy threshold above: it is a real contractual term, it is undisclosed, and asking for it in writing is a legitimate procurement question with a checkable answer.

So what should you actually buy?

Work out whether you need phone numbers first, because it changes the answer more than budget does. Then run the test before the contract, not after.

• Email only, any volume: Hunter or Prospeo at roughly 2.5 cents an email. Hunter's free tier is 50 credits a month and Prospeo's is 100, both with export, so both can be tested at zero cost before a card goes in.
• Reps who dial: LeadMagic at 15 cents a complete contact, on the strength of the 5x multiplier rather than the headline price. It has no free tier and publishes a 14-day refund on a first payment, which is a better test route than most free tiers because export is unrestricted.
• Then check the crossover before you commit. Prospeo only has to be 63 percent better than LeadMagic on usable rows to become the cheaper option, which one morning of testing will settle. Nothing further down the table needs checking, because no vendor is 300 percent better than another on the same list.
• Hard lists where single sources fail: a waterfall like FullEnrich or BetterContact, but only below about a 65 percent existing hit rate. Above 70 percent the extra contacts cost 56 cents against 15 to buy one outright, and you should spend the money on verification or on more accounts.
• Verification, large list, mostly big-company domains: NeverBounce, which is meaningfully cheaper at scale and whose billing ambiguity costs you little if your list is light on catch-alls.
• Verification, catch-all-heavy list or a forecastable invoice: Clearout or MillionVerifier. Clearout publishes the full rule and its credits never expire; MillionVerifier states it charges for neither unknowns nor catch-alls, which is the most buyer-favourable published rule of the seven.
• Before any of it: run the 200-record protocol. It costs about $52 and half a day, and against an annual data contract it is the cheapest thing you will do all quarter.

What does one contact cost? The answer

Between 15 and 83 cents at entry pricing for a contact you can email and call, from the six vendors who publish enough to compute it, at prices read in August 2026. That is the answer to the title and it is the numerator.

The real figure is that number divided by the share of rows that come back and turn out to be right, which on ordinary B2B lists puts the same six vendors somewhere between 19 cents and four dollars. Nobody publishes the divisor. Eleven vendors publish an accuracy percentage and none of them publishes a method. The one company that contractually guarantees accuracy told the SEC it does so without telling anyone the threshold.

What the arithmetic does settle is which comparisons are worth arguing about. LeadMagic's 63 percent price advantage over Prospeo is inside the range a single morning of testing can overturn, so do not decide it on price. Its 450 percent advantage over BetterContact is not, so on like-for-like billing you can decide that one on price and move on.

Which leaves one thing to do this week. Pull 200 records from the worst part of your ICP, run them through Prospeo's free 100 and LeadMagic's refund window, verify the results through Clearout's free 100 and Emailable's free 250, and send to them. You will have your own denominator by Friday, and it is the only number on this subject that will predict your own bill.