Validate Before You Enrich: How to QA a 20-Profile Apollo Sample

The most expensive Apollo mistake usually happens before anyone buys an email address. It happens when a search is treated as a finished list.

The result count looks healthy, the titles sound close enough, and the geography label appears correct. The team exports thousands of records, enriches them, and only then notices that the search mixed company headquarters with person location, included adjacent roles, admitted excluded industries, or returned subsidiaries that do not match the offer.

Enrichment did not create the targeting problem. It simply made the problem more expensive.

Why Apollo search results need an acceptance test

Apollo's own filter documentation explains that filters work together and that result counts can change as records, saves, enrichment, and scoring change. It also separates Total, Net New, and Saved results. That means a screenshot of a large count is not a stable definition of the list.

The most important distinction is geography. Apollo lets an operator search by an individual's location or by account headquarters. Those answer different business questions. A US-headquartered company can employ a person in another market, and a person in the target market can work for a company headquartered elsewhere.

Title logic can drift too. Similar-title matching may be useful for discovery, but it can also admit people whose day-to-day responsibility is outside the buying committee. Industry categories, company size, ownership structure, and parent-subsidiary relationships create similar edge cases.

This is why the ICP-to-Apollo filter workflow should produce two things: the search itself and a testable targeting contract.

Step 1: freeze the targeting contract

Before reviewing profiles, write down exactly what counts as a match. Do not rely on the operator remembering the sales call.

person location:
account headquarters:
included industries:
excluded industries:
company size:
included titles:
excluded titles:
seniority:
required company evidence:
required contact fields:
duplicate definition:
hard-fail conditions:
human-review conditions:

Hard constraints should be binary. If the campaign requires decision-makers physically located in the United States, a company-HQ match cannot substitute for a person-location match. If agencies are excluded, an agency does not become acceptable because its founder has the right title.

Soft signals can be scored separately. Website evidence depth, likely buying urgency, and contactability may tolerate uncertainty, but that uncertainty should remain visible as CHECK.

Step 2: choose a difficult 20-profile sample

Do not review only the first 20 polished profiles. That checks whether Apollo can show obvious matches. It does not test the filter edges.

A useful sample includes:

• profiles from early and later result pages
• exact titles and similar titles
• small, medium, and upper-bound companies
• people whose location differs from company headquarters
• subsidiaries, portfolio companies, and parent organizations
• borderline industries and business models
• records with and without the evidence needed for personalization
• potential duplicates already present in the CRM or another source file

The point is controlled friction. If every sampled profile is easy, the test is too comfortable.

Step 3: score every profile against the same gates

For each profile, preserve the Apollo URL, company URL, decision, reason, and reviewer. If external evidence changes the decision, record that source too. This creates a compact evidence ledger instead of a spreadsheet full of unexplained colors.

We use the same discipline in larger list operations: stable lead IDs, preserved source exports, explicit exclusions, and a separation between evidence-backed approval and downstream delivery. A large raw export remains a source artifact until the ICP and campaign are approved. Volume alone does not authorize cleaning, enrichment, or upload.

Step 4: calculate projected enrichment waste

Suppose 14 of 20 sampled profiles pass, two require review, and four fail. The observed hard-fail rate is 20 percent. On a 5,000-record universe, enriching first could expose roughly 1,000 obviously unsuitable rows to paid lookups before softer review even begins.

That projection is directional, not statistically precise. It is still useful for an operational decision:

sample size = 20
hard passes = 14
checks = 2
hard fails = 4

observed hard-fail rate = 4 / 20 = 20%
projected hard-fail rows = universe size x 20%
projected review rows = universe size x 10%

Combine that with the current Apollo action and workspace rules. Apollo documents different credit categories for data requests, enrichment, exports, phone data, and AI research depending on plan and configuration. Review the current credit estimate and usage history rather than publishing one universal cost assumption.

Step 5: define the approval threshold before seeing the result

A practical default is 20 out of 20 on hard constraints such as geography, exclusions, and identity. Softer signals can use a declared threshold, but the threshold must be chosen before the team sees whether the current search passes.

Reject or revise the search when:

• any excluded category appears repeatedly
• person location and account location were confused
• similar titles produce a different buying role
• parent and subsidiary records create duplicate account coverage
• the sample lacks enough evidence to support fit decisions
• the projected usable yield no longer justifies the enrichment cost

If the search fails, change one filter family at a time and run a fresh sample. Keep the old filter state and decision so the team can explain why the final search differs from the original request.

What happens after the sample passes?

1. Save the exact search and timestamp the result count.
2. Export or save only the approved population.
3. Preserve the original source fields before enrichment.
4. Run enrichment through a controlled provider waterfall.
5. Separate matched, duplicate, not-found, and unverifiable outcomes.
6. Apply suppression, CRM-history, and campaign-approval gates.
7. Read the destination back after any CRM or sequencer write.

The enrichment waterfall reduces duplicate paid lookups after this point. It cannot rescue a bad target universe. The sample gate must come first.

Frequently asked questions

Is a 20-profile sample statistically representative?

Not necessarily. It is an operational acceptance test designed to expose obvious filter failure and edge cases before scale. Use a larger random sample when you need a formal confidence estimate.

Should I sample before or after Apollo enrichment?

Sample before paid enrichment whenever the profile and company fields are sufficient to evaluate targeting. Enrich only the approved population, then run a second QA pass on contactability and newly added fields.

Why do Apollo saved-search counts change?

Apollo says its data changes over time as people and companies are added or removed, contacts move between Net New and Saved, and enrichment or scoring updates affect matching. Timestamp the filter state and count used for approval.

What is the biggest Apollo geography mistake?

Using company headquarters when the campaign requires the person's location, or doing the reverse. Define the intended geography field explicitly and test records where the two differ.

Does a verified email mean the lead is campaign-ready?

No. Email status does not prove ICP fit, suppression clearance, sender readiness, legal suitability, or inbox placement. It is one field inside a larger release decision.

AnswerThePublic research note

The English and United States AnswerThePublic report for Apollo lead list validation showed search volume 0, CPC 0, no organic keyword results, and no People Also Ask results on July 29, 2026. We kept the wording because it precisely describes the workflow, not because the tool showed demand. The article also uses natural variants such as Apollo sample QA, Apollo search filters, validate leads before enrichment, and Apollo enrichment credits.

Sources