Clay in the Era of MoltSets and BlitzAPI: Retain or Retire?

Retire Clay as the first step of an enrichment waterfall. Keep it for two jobs where a logged test showed it adding results: checking whether someone still works at a company, and finding a second email source for the people the first source missed. That is how our own waterfall runs as of October 2026. Employment check: BlitzAPI, then Clay. Work email: MoltSets, then Clay on the misses. Clay is last in both.

Between July and early October Clay went from running one large one-off job on its own to being the last step. Your lists will differ, so run the same test on 200 of your own rows before you change anything.

Where Clay sat in our stack, and where it sits now

For one large one-off job in late July and early August, Clay ran the enrichment. Its credit balance fell by 28,302.5 in about ten days. The full ledger comes below. Then the order changed. By 26 September the employment check ran BlitzAPI first and Clay second, and on 28 September the email test ran MoltSets first, with Clay on the misses. Since October the written order is BlitzAPI then Clay for employment. MoltSets then Clay for email.

These four days are not like for like, and the figures come from balance readings in our ledger. The 28 September figure was a deliberate side-by-side test, and 7 October was 14 people, all answered by BlitzAPI. Read the table as a log of what Clay was asked to do, not a trend.

Four days is a small sample. And the July ledger sits on Clay's legacy plan and the four days above sit on the plan we use now, so don't compare the totals.

Retire: first-pass email

Start with first-pass email. On the same 821 profiles, MoltSets alone returned 356 usable business addresses (43%) on 28 September and 411 (50%) on 22 October. Usable means the address sits on the person's own company domain, or on a known alias of it. Addresses on another company's domain fell from 91 to 40. But coverage of any address stayed flat at 54% to 55%. Ten B2B tech companies, mostly US-based. One source. And "usable" tests the domain, not whether the mailbox belongs to that person.

MoltSets sells that plan at a flat $27 a month, with a rate limit of 1,000 enrichment records per five hours (its pricing page, as of October 2026). Clay meters by the credit. On 28 September it cost 573.5 credits to look for addresses for the 461 people MoltSets had missed (a logged balance reading). The logs don't let us reconcile addresses found against addresses usable, so we print no cost per address, and we leave out a Clay-only hit rate because the figure we had doesn't reproduce from our archive.

The units differ, since one is a subscription and the other is a meter.

Employment went a different way. In a 28 September test of who still worked where we thought, BlitzAPI answered 14 of 15 correctly, with one no-record. MoltSets managed 5 of 11, and we have not re-run it since. Small samples, one day. But enough to put BlitzAPI at the front of the employment check, and Clay's own 12 of 12 comes below.

BlitzAPI also publishes a benchmark that ranks BlitzAPI first. It was run by BlitzAPI, one of the six providers measured, on 3,445 LinkedIn profiles on 17 September 2026. Accurate records per 100 contacts: BlitzAPI 82.4, Apollo 70.0, Prospeo 52.7, QuickEnrich 43.8, MoltSets 43.7, GetLeads.io 31.9. Clay is not in it. Read it as a vendor-run claim to reproduce on your own list. It also measures something different from our test, accurate records for a profile, not usable business addresses.

Retain: the employment fallback and the second email source

Clay was right on all 12 people in a 28 September employment test. Small sample. Twelve people prove little, which is why a human opens the profile when the automated steps disagree or come back empty. But it's enough to keep Clay as the step after BlitzAPI. And we only feed it profile URLs, because an email-only lookup has matched the wrong person.

The per-account cost is small. On 2 October, four of the five accounts we ran logged 6.0, 11.3, 14.7 and 56.4 Clay credits by stage (employment, email, second source; the fifth, a pilot, is in the 95.8 day total but not in the archive's stage log), and gate runs cost 0.5 to 6.0 credits, from one candidate check to a whole account's gate. The balance fell by less than the logged stage sums on three of the four accounts, and we don't know why.

The second email source is the stronger case. In September, MoltSets first and Clay on the misses reached 583 usable addresses out of 821 (71.0%). Clay's part was 227: it returned 235 new addresses, and 227 survived our domain and owner filters. And Clay wasn't re-run on 22 October. MoltSets alone had reached 411 by then, so we can't say how many of the 227 would still be new today. But the 573.5 credits above came from the same test, and the logs don't tie them to the 227.

There are two smaller reasons to keep it. If you bring your own API keys, Clay charges one Action per run and no data credit (more in the pricing section). And its tables are built for steps a teammate can edit without writing code. Or you can run those steps in an agent instead. That's a separate decision, covered in Claygent vs Claude Code for GTM research.

What the meter read: 2,443 contacts on a legacy plan

Between 28 July and 6 August 2026 we ran a large enrichment job through Clay's legacy plan, a single credit pool that started at 64,237.5. The routine bundled an email and a phone lookup, and Clay quoted it at a ceiling of 18.2 credits a contact. We ran it in stages and read the balance before and after each one.

The total counts the pilot once. Then 470 plus 6 plus 400 plus 1,567 is 2,443 contacts. Credits add up to 28,292.5, and 28,292.5 divided by 2,443 is 11.58. But the balance itself fell from 64,237.5 to 35,935, a drop of 28,302.5. That leaves 10.0 credits not tied to any run.

Against the quote, 18.2 minus 11.58 is 6.62, and 6.62 divided by 18.2 is 36.4%. Spend stayed about 36% under the quote. And the three bulk runs landed within 7.0% of each other, from 11.15 to 11.93 a contact. The worst case for the 400-contact test was 400 times 18.2, or 7,280 credits. It cost 4,772.

One more reading from that job. The live meter climbed to 5,669 during the 400-contact test, then settled at 4,772. Our reading is that Clay reserves credits when a source call goes out and releases them when the call comes back empty. A meter watched mid-run overstates the final charge.

But legacy credits and current credits are different units. These 11.58 credits belong to one routine on the old plan, and the current plan meters in data credits and actions, so we don't compare the two.

Clay pricing in October 2026: plans, credits and actions

Clay's docs and pricing page list the prices below. We read both on 27 October 2026, and the figures are list prices on monthly billing.

Launch and Growth cap a table at 50,000 rows, and the free plan caps it at 200. But the pricing page cards show $167 and $446. Those match the annual-billing toggle at 10% off ($185 times 0.9 is $166.50). A low number on the page usually means the annual toggle is on, so check which one is set.

Two meters sit behind the price. Actions measure orchestration: one per enrichment or execution, which Clay's docs put at a few tenths of a penny each. Data credits pay for third-party data and AI, at "a few pennies" each in the same docs. Bring your own API keys and each run counts as one Action with no data credit. A waterfall charges only the source that finds a match. And an enrichment that returns nothing costs neither credits nor Actions. We cover the split in Clay data credits versus actions and the API side in Clay API explained.

Rollover differs for the two. Actions reset each billing cycle. And on monthly plans, the docs and the pricing FAQ say unused credits accumulate up to twice the monthly amount, while the comparison table on the pricing page lists monthly rollover as up to one month's credits. For annual plans the docs say up to 15% of annual credits, if you renew on the same or a higher tier, while the pricing FAQ ties the 15% to Enterprise and the plan comparison on the same page lists it under annual Launch and Growth. The docs, the pricing FAQ and the plan comparison word it three ways. Confirm with Clay before you plan around it.

Clay prices credits through an interactive selector on its pricing page, so there is no single published rate per credit to quote. Comparing Clay with other enrichment sources on price? FullEnrich vs Clay and Apollo vs ZoomInfo vs Clay compare them head to head. Weighing an intent platform rather than a data API is a different decision, and Clay vs 6sense covers it. The figures above are dated, so check the page before you buy.

Waterfall enrichment without Clay first

A waterfall enrichment asks one source for an answer and sends only the misses to the next, stopping when a valid result comes back. And Clay's docs describe its work-email waterfall in the same terms. And order decides cost, because the first source's hits never reach the paid steps.

BlitzAPI's September benchmark (vendor-run, 17 September, 3,445 profiles) shows how fast the later steps shrink. Querying a second source on the misses lifted accurate records from 82.4 to 86.9 per 100 contacts. "A third adds 0.3." The chain runs Apollo second and MoltSets third, a sorting rule rather than a recommendation, per the page. The best possible result with all six is 91.2, by the page's own figure. Apollo is not in our stack, so our order differs. But the shape holds: in this benchmark the second step supplied nearly all of the chain's gain.

For 10,000 contacts, the arithmetic from published caps and our own logged rates comes to this.

BlitzAPI counts records returned, MoltSets counts enrichment records inside a time window, and Clay counts credits that depend on which sources you order. Clay's docs list Growth credits from 6,000 to 100,000 a month, so 86,800 credits in one month would sit near the top of that range. And that's one reason we print credits and not a dollar total.

BlitzAPI's $99 Start plan looks very interesting to us and is under consideration. And Start includes every endpoint, verified work emails and US phone numbers, and a job of this size would take 15% of its monthly cap. The $99 is a flat monthly price, per the pricing page. But MoltSets' time-windowed limit decides how long a run takes.

For the cost mechanics behind all three, see the enrichment waterfall that bleeds money, our waterfall teardown and does the Clay API lower waterfall cost. Tool profiles: BlitzAPI, MoltSets, Clay and the codified stack.

How to decide: retain or retire

Decide job by job, and test on 200 of your own rows first.