How many publishers do you need to test before scaling?

Somewhere between 15 and 40, depending on the vertical. That's the honest answer, and I know it's not the clean number people want when they ask me this at conferences. But insurance lead gen doesn't reward clean numbers. It rewards people who understand why the range exists and where their campaign falls inside it.

I've had this conversation more times than I can count with agencies who want to scale after testing five publishers. Five is not a test. Five is a guess with a spreadsheet attached.

Why the publisher count depends on the vertical

Here's the thing: Medicare and final expense aren't the same animal, and treating them the same during testing is how buyers burn budget without learning anything.

Medicare Advantage and Medicare Supplement campaigns typically need 20 to 40 or more publishers in the testing pool. Why so many? AEP (October 15 through December 7) and OEP (January 1 through March 31) create volume spikes that completely rearrange the leaderboard. A publisher that looks mediocre in August can turn into your best source in November. One that crushes it in Q1 can go quiet the rest of the year. Test during a slow season with a small pool, and you're scaling on data that won't hold up when volume actually matters.

Final expense and life insurance, by contrast, tend to need fewer publishers in the initial round, usually somewhere in the 10 to 20 range. Call center conversion patterns in these verticals stay more consistent across traffic sources. A senior looking into final expense behaves fairly predictably whether they came from a native ad or an SEO landing page. That consistency means you can draw conclusions faster with a smaller pool.

Auto insurance sits somewhere in between. Health insurance u65 tends to run closer to the Medicare end of the spectrum because of how much variance shows up across traffic sources. In practice, u65 leads can swing wildly in quality depending on whether the traffic came from paid social, co-registration, or organic search, even when the raw volume numbers look similar on a dashboard.

Match your publisher count to the vertical's natural variance. Not to a round number that sounds good in a meeting.

The spend threshold that actually matters

A lot of buyers make scaling decisions off 10 clicks and a bad feeling, or 10 clicks and a good feeling, which is honestly worse because it convinces them to scale. Neither tells you anything.

Most experienced media buyers won't draw conclusions on a publisher until it's hit somewhere between $500 and $2,000 in spend, or 50 to 100 clicks or leads, whichever comes first. Below that threshold, conversion rates bounce around so much you're basically reading tea leaves. A publisher converting at 2% on 20 leads might be converting at 8% by lead 80. You don't know yet. Pretending you do is how good sources get killed early and bad sources get funded too long.

I've watched agencies cut a publisher after $300 in spend because the first dozen leads looked rough, only to find out later that publisher was one of their top three performers once it had real volume behind it. The math on small samples lies to you constantly. It's not that the data is wrong. There just isn't enough of it yet.

For call center partners, especially common in final expense and Medicare, the testing window needs to stretch even further. A 90-day minimum per publisher is standard for a reason: agent performance variability is a separate variable from lead quality, and 90 days gives you enough calls to average out both a good agent day and a bad one. Cut a test at 30 days and you might just be measuring which agent happened to be on shift.

Segment by traffic type, not just by publisher name

This is the part buyers skip. And it costs them.

Two publishers can drive the exact same volume, look identical on a report, and still produce completely different outcomes once you dig into how that traffic was generated. SEO traffic behaves differently than paid social. Native ad traffic behaves differently than co-registration. A publisher blending all four channels under one name in your reporting is hiding the real story from you.

You need to know, publisher by publisher, what channel the volume actually comes from. Co-reg traffic in particular tends to produce leads that look fine on volume and terrible on intent, and it's also where a lot of compliance risk lives. If you're not segmenting by traffic type during testing, you're not really testing publishers. You're testing an average of several different things wearing one publisher's name tag.

This matters even more given the compliance environment insurance lead gen operates in right now. TCPA litigation risk is real money, not theoretical risk, and most serious buyers run every publisher's leads through a platform like Jornaya or TrustedForm during testing to validate that consent was actually captured the way it's claimed. A publisher that won't cooperate with compliance validation during the test phase is telling you something about how they'll behave once real money is on the line. Don't ignore that signal because the CPA looks attractive.

CPA ranges vary enough by vertical that comparing raw numbers across categories is pointless anyway. Auto insurance leads commonly run $5 to $30 per lead. Medicare leads, particularly Medicare Advantage, can run $20 to $60 or more depending on exclusivity and how much data verification the buyer requires. Testing a Medicare publisher against an auto benchmark in your head means you're already comparing apples to a different fruit entirely.

State-level regulation adds another layer. NAIC guidance and individual state DOI rules affect how leads can be tested and distributed, and Medicare products carry additional CMS marketing guidelines on top of that. A testing protocol that works fine in one state can put you sideways in another, so your publisher pool and your compliance checks need to account for state variation, not just vertical variation.

A publisher isn't one thing. It's a mix of traffic sources, and you have to test the mix, not the label.

If you're more interested in building your own inbound call volume instead of buying it from publishers, that's a different skill set entirely, and it's worth learning properly rather than piecing it together from forum posts. I wrote The Pay Per Call Revolution for exactly that reason, and there's a companion workbook that walks through building the system step by step.

FAQ

How long should I test a new publisher before deciding to scale or cut them? At minimum, until they hit $500 to $2,000 in spend or 50 to 100 leads. For call center partnerships, plan on a 90-day window to separate agent variability from actual lead quality.

Do I need to retest publishers seasonally for Medicare campaigns? Yes. AEP and OEP volume spikes shift publisher performance enough that pre-season testing data can become unreliable once real volume hits in October through December and again in January through March.

Is a smaller publisher pool ever acceptable? For final expense and life insurance, a pool of 10 to 20 is usually enough given more consistent conversion patterns. Health insurance u65 and Medicare need larger pools, 20 to 40-plus, because of higher variance.

What's the biggest mistake buyers make when testing publishers? Judging performance before hitting a real sample size, and failing to segment results by traffic type. A publisher blending SEO, native, and co-reg traffic can hide serious quality and compliance problems inside an average that looks fine on paper.