Apollo filters that inflate your TAM, and the ones that hold up

A narrow, industry-specific title search in Apollo returned 679 people in the United States. The same intent, with broader titles and one default setting left on, returned 91,649.

Neither number is an error message. Both look like a market. One of them is a market; the other is a campaign full of the wrong people that reads like progress right up until the replies come in.

That’s the problem with Apollo’s match count. It is the first number anyone sees, it drives the size of the list, the number of mailboxes, and the credit budget, and several filters inflate it without telling you. Here are the ones we have caught on real searches, and the check for each.

The filters that inflate the count

1. Similar titles, on by default

Apollo’s people search has an “include similar titles” setting, and it defaults to on. With a specific title it barely matters. With broad titles like “marketing manager” or “events manager”, it is a volume multiplier: it is what turned 679 into 91,649.

The check: run the search once with similar titles off and only the strict core titles. That number is your real base. Widen from there on purpose, one title at a time, and watch what each addition brings in.

2. “Founder” matches founding employees

A title search for founder also returns Founding Engineer, Founding Designer, Founding GTM, Founding Recruiter, and Founding Member. None of them buys outbound, and in our runs they were 9 to 14% of the list. Nobody notices until a founding engineer replies asking why you are pitching them.

The same trap sits in seniority. The C-suite bucket sweeps in Chief People Officer, Chief Business Officer, and other roles that may have nothing to do with your offer.

The check: scroll the title column on the first few pages of results before you save anything. Every near miss you spot goes into the job title exclusions, so it stays out of every search after this one.

3. Industries are broad, and keywords are loose

Apollo’s Industry filter is a fixed list, and each company carries one industry, the one it picked for its own profile. There is no SaaS industry and no B2B industry. A software company might sit under information technology & services, computer software, internet, or management consulting, depending on what someone chose when they set up the company page. Information technology & services alone is a huge bucket: in our own Apollo exports it is more than half of all rows.

So people reach for company keywords, which match the company’s name, keywords, or description. That gets closer, but a broad keyword like “software” or “technology” also matches every agency, consultancy, and reseller that mentions the word.

The check: after the search, open 20 or 30 companies from the results and look at what they actually sell. If a big share are agencies, consultancies, or hospitals, tighten the keywords, set them to match all rather than any, or exclude those industries with “is not any of”.

4. A funding filter without a stage lets acquisitions through

“Recently funded” is one of the better timing signals Apollo offers, but a funding date range on its own also matches acquisitions. Add the funding stage every time you filter on funding, or the list fills with companies that were bought rather than backed.

Filters that hold up for SaaS and services

For a B2B seller whose buyer is the founder or CEO of an 11 to 200 person SaaS or services company in the US, this combination has held up:

Then add one timing signal, not five. The ones worth testing first:

• Hiring sales roles. Companies with open SDR or account executive postings have decided outbound matters and are about to pay a salary to solve it.

• Headcount growth over the last six or twelve months.

• Recently funded, always with the funding stage set.

• Sales team size, for example at least five people in sales, where your plan allows the department filter.

Each signal is a separate list and a separate test, so you can see which one actually produces replies.

Check before you spend: the five-minute routine

1. Run the strict search with similar titles off, and note the count.

2. Widen it one change at a time, and note what each change adds.

3. Scroll the first few pages. Add near-miss titles to the exclusions, and exclude the industries that drifted in.

4. Save only what survived to a list, and only then spend credits on emails.

The rule underneath all of it: a filter expresses your intent. It does not guarantee the result. Read what came back before you pay for it.

If an agent runs the search for you

More teams now run Apollo through Claude or ChatGPT, over the API or Apollo’s MCP. An agent can check things no person would do by hand, and it should, because it also pulls lists far faster than anyone would scroll them:

• Read the count first. The match count comes back as total_entries before anything is pulled. Anything over 50,000 (500 pages of 100) is too broad, so add filters rather than slicing it into pieces.

• Count titles and industries across the whole pull, not a sample. The industry field can come back empty in search results, but the NAICS and SIC codes are filled, and they are enough. On one real list of 3,238 people, 43% sat in professional and technical services and 19% in software, with a tail of education and health care that was cut before a credit was spent.

• Dedupe before enriching. Pages have overlapped on our pulls: 33 pages returned 3,238 rows holding 2,933 unique people, a 9% duplication rate. Enrichment charges for every row, so every duplicate is a credit spent twice.

• Treat zero as a question, not an answer. Technology filters use slugs that are not always obvious (Apollo itself is apollo, and apollo_io returns zero), an unknown department key returns nothing, and some filters are plan-gated. If a search that should have results returns none, check the filter values before concluding the market is empty.

Run it yourself, or have it built

The agent checks above are built into the list builder and the list quality scorecard in Apollo Operator, our free, open-source toolkit for Apollo. The ICP builder prompt and the lead list prompt run the same method in any chat. If you want the list built, verified, and handed over as files you keep, that is Data & Consulting; if you want it run inside your own Apollo account, that is Apollo Managed.