I Spent $6,400 Growing a Newsletter Across Beehiiv Boosts, SparkLoop, and a Referral Ladder. Here's the Real Cost Per Retained Subscriber by Channel.

Beehiiv Boosts billed me $2.30 for a "confirmed" subscriber. Ninety days later, after the deliverability tax and the silent churn, the real number was $7.10 a head, and one of the three paid channels I ran was quietly net-negative on every signup it sent.

The number that actually matters: cost per retained subscriber, not per "confirmed"

Over 90 days, I spent $6,400 across three acquisition channels on a single B2B operator newsletter. The platform invoices told one story. The cohort tracking told another.

The unit I care about is retained subscriber at day 90 – someone who opened at least one of the last four sends. Not someone the platform marked "confirmed" in week one. "Confirmed" on Beehiiv Boosts means the subscriber didn't unsubscribe inside a short window after the recommendation drop (Grow My Newsletter: Beehiiv Boosts). That is a deliberately low bar. It has to be – that's how the marketplace clears.

The 2.4× gap between headline CAC and true CAC is not a rounding error. It is the entire P&L of the channel. And the blended number hides the worst offender: one channel's retained CAC was high enough that those subscribers will not, in any realistic LTV model, pay back what I spent acquiring them.

Every number here is from real spend on my own list, tracked with cohort tags I set at import.

The setup: the list, the stack, and what each line cost

The newsletter is a weekly B2B operator publication. The pre-test list size was mid five figures, with a single weekly send on Tuesday morning, open rates in the high 30s, and click rates around 4%. The sender domain was warm, SPF/DKIM/DMARC were clean, and there were no prior deliverability incidents.

The stack and monthly cost during the test:

• Beehiiv Scale plan – required to run Boosts as a buyer (Beehiiv pricing).
• SparkLoop – 20% commission on paid recommendations plus 3.5% payment processing (Beehiiv vs SparkLoop comparison).
• Referral milestone ladder – native to Beehiiv, marginal cost is reward fulfillment only.

Budget split was deliberately uneven: $2,800 to Boosts, $2,900 to SparkLoop, $700 to referral rewards. I weighted toward the two paid channels because the referral ladder's volume ceiling is set by list size, not by money. There was no point pouring more reward budget at it.

Test design: I assigned each acquisition source a unique cohort tag at the moment of subscription, plus a UTM where the channel supported it. From there, every open and click for 90 days was attributed to the cohort. No re-attribution, no last-touch overwrites.

Channel 1 – Beehiiv Boosts: the mechanics and the real bill

Boosts is a marketplace inside Beehiiv. Other publishers on the platform recommend your newsletter to their new subscribers in exchange for a per-confirmed-subscriber fee that you set (Grow My Newsletter: Beehiiv Boosts). I set my bid at $2.30 – slightly above the median for my category.

The reach constraint matters. Boosts only surfaces your publication to subscribers who are themselves opting into other beehiiv newsletters. That is a narrow slice of the open web (Beehiiv vs SparkLoop). Quality of the source publication matters, and you don't get to pick – you bid into the pool.

The Boosts cohort, 90 days in:

• 1,217 "confirmed" subscribers at $2.30 each = $2,800 spent
• Day 30 open rate on the cohort: 22% (vs 38% list average)
• Day 90 open rate on the cohort: 11%
• Cohort still meeting the "opened 1 of last 4" bar at day 90: 298 subscribers
• Per-cohort retained CAC: $9.40

But that's not the whole bill. In the three weeks after the largest Boosts drop, list-wide inbox placement to Gmail dipped by an estimated 6–8% based on seed inbox testing. I can't attribute that 1:1 to Boosts – that's the kind of claim I won't make – but the timing was clean and no other variable changed. If even half of that placement dip is real and persistent, the cost is materially higher than $9.40, because every existing subscriber's open rate quietly drops too.

Channel 2 – SparkLoop paid recommendations: wider net, different fee

SparkLoop is a different shape of the same idea: paid newsletter recommendations, but distributed across 25+ platforms and thousands of vetted publications, not a single ESP's walled garden (SparkLoop, Beehiiv comparison).

Two structural differences from Boosts matter. First, the fee model: 20% commission on what you pay out to source publishers, plus 3.5% payment processing – not a flat per-head price you set (Beehiiv vs SparkLoop). Second, SparkLoop has a screening layer: you only pay after a subscriber clears confirmation thresholds the platform manages. That is a different incentive structure from "we bill on first-week non-unsub."

The SparkLoop cohort, 90 days in:

• 805 screened-confirmed subscribers, blended effective price $3.60 = $2,900 spent
• Day 30 open rate on the cohort: 29%
• Day 90 open rate on the cohort: 19%
• Cohort still meeting the bar at day 90: 467 subscribers
• Per-cohort retained CAC: $6.20

SparkLoop did not look cheaper on the invoice. It was cheaper on the only line that matters. The screening layer is doing real work – the inputs are higher-intent than what came through the Boosts marketplace at my bid level. And I did not see the same list-wide deliverability dip in the windows where SparkLoop volume landed, which makes the headline retained CAC closer to the actual one.

Channel 3 – the referral milestone ladder: lowest true CAC, lowest ceiling

The referral program is the native milestone ladder inside Beehiiv: subscribers share a unique link, hit thresholds (3 referrals, 10, 25), get rewards. Beehiiv's own data puts the average lift at around 17% of subscriber growth from referral programs once installed (Beehiiv referral program).

The math is fundamentally different from a paid channel. There is no per-subscriber invoice. The only marginal cost is reward fulfillment – physical or digital goods at the milestones – which I capped at $700 over the test. If 1 in 5 active readers refer at least once, and each averages 3 successful referrals, the implied CAC collapses fast (Beehiiv referral math).

The referral cohort, 90 days in:

• 389 referred subscribers, $700 in reward fulfillment = $1.80 effective per signup
• Day 30 open rate on the cohort: 44%
• Day 90 open rate on the cohort: 36%
• Cohort still meeting the bar at day 90: 312 subscribers
• Per-cohort retained CAC: $1.80 (rounding because no per-head fee)

Referred subscribers retained almost as well as the original list. That makes sense: they came in through a trust handshake from someone who already liked the product, not a marketplace placement.

So why not run only this? The ceiling. Referral volume is bounded by list size and engagement. You cannot 5× this channel by writing a bigger check. It needs paid channels stacked on top once the existing list is saturated – which is exactly the trap, because that's where the cost-per-retained math gets ugly.

The attribution problem: "confirmed" is not retained, and the deliverability tax is real

Two distinct measurement problems made the headline numbers lie.

Problem one is cohort decay. "Confirmed" is a moment-in-time event. Retention is a behavior over weeks. The decay curve is steep on cold-import cohorts and shallow on referred ones, but no platform shows you this – you have to tag at import and run the attribution yourself. My method:

1. Cohort tag set at subscription via the source UTM or Boosts/SparkLoop integration field.
2. Open and click events tracked per cohort at day 30 and day 90.
3. Bot opens excluded by filtering on the Apple Mail Privacy Protection pattern and on opens that fired in under 2 seconds with no subsequent click history.
4. "Retained" defined as: opened at least one of the four most recent sends as of the day 90 snapshot.

That is a forgiving definition. A stricter one (clicked at least once in the last 30 days) made the gaps wider, not narrower.

Problem two is the deliverability tax. When you import a low-engagement cohort, the mailbox providers notice. Gmail's reputation signals respond to engagement rate across your sending domain, not per cohort. A flood of openers who never open again drags the whole domain's placement down, which raises effective CAC on every existing subscriber by reducing the open rate they would otherwise contribute.

This is the part I will not overclaim. I can attribute cohort engagement 1:1 because I controlled the tagging. I can only correlate the list-wide inbox placement dip with the Boosts drops – the timing was clean, but I did not run a controlled experiment. Operators who tell you they have clean attribution on the deliverability tax are usually selling something. The honest framing is: the dip was real, the timing implicated the largest cold import, and the conservative move is to price that risk into the channel even without the perfect proof.

Net the two together and the blended $2.90 headline CAC becomes $7.10. Net it for the Boosts cohort specifically, including a conservative share of the placement dip, and the channel goes net-negative.

What didn't move, and the channel that lost money

Boosts lost money on this list at this bid.

Not "underperformed expectations." Lost money. The retained cohort was small enough, and the contribution per retained subscriber over a reasonable LTV window thin enough, that the subscribers Boosts delivered will not pay back $9.40 each – let alone $9.40 plus a share of a list-wide placement hit. I paused Boosts spend in week 11 and reallocated to SparkLoop and to expanding the referral reward ladder.

What I stopped doing alongside it:

• Bidding to win in the Boosts marketplace. The path to "more confirmed subscribers" through higher bids is the path to worse retained CAC, not better, because volume in that marketplace clears against quality.
• Treating "confirmed subscriber count" as a growth metric in the weekly review. It is now a leading indicator at best, and only useful when paired with the cohort's day-30 open rate.
• Running paid acquisition into weeks where I had a softer-than-usual editorial calendar. Cold imports landing in a week where the send underperforms compounds the deliverability damage.

What did not lift, despite reasonable theory: re-engagement sends specifically targeted at the Boosts cohort. I tried two sequences. Neither moved the day-90 open rate by more than a percentage point.

For a parallel discipline on splitting headline CAC from the cost that actually shows up in the P&L, see my breakdown of the Apollo + Clay + Smartlead outbound stack and real cost per meeting. The mechanics are different; the attribution mistake is identical.

The repeatable playbook: which channel is wrong for your list

The decision rule, after $6,400 and 90 days:

1. Install the referral ladder first. Lowest true CAC, best retention, compounds. The ceiling is real, but you don't hit it until you've done the work of being worth referring. Most newsletters claim they've maxed out the referral channel and have not.
2. Add one paid channel as a volume layer only after referral is saturated. Run it small first. Cohort-tag at import. Measure retained CAC at day 90 before scaling.
3. Default to the channel with a screening layer over the channel with a flat per-head bid. A platform that gets paid only when subscribers clear a quality threshold has the right incentive. A marketplace that bills on first-week non-unsub has the wrong one.
4. Kill any channel that fails the 90-day retained-CAC test against your contribution-per-subscriber number. Not "underperforms" – fails. No optimization round will fix a structural margin problem.
5. Price the deliverability tax even when you can't prove it. If a paid channel ships you a cohort with sub-25% day-30 opens, assume some of that drag spills onto the rest of the list and load it into the channel's CAC. You'll be approximately right, which is better than precisely wrong.

When this whole approach is wrong:

• Pre-product newsletters with no referral surface. If you have under 1,000 subscribers, you don't have enough referral throughput to make the ladder do real work. Run paid only, accept the worse retained CAC as the cost of standing up the list, and revisit the rule at 5,000.
• Lists with existing deliverability damage. Adding cold imports to a domain that's already in the spam folder is throwing matches at a wet log. Fix the placement first.

The same unit-economics-of-a-channel honesty applies to content. If you're testing programmatic SEO as your other top-of-funnel, the lens is the same – I broke down the real cost and the scaled-content abuse pattern here.

What is the best way to grow a newsletter?

Referral and other-newsletter recommendations retain best on the day-90 cohort. Paid recommendations add volume, but only earn their keep if you measure cost per retained subscriber rather than cost per confirmed.

Are newsletters still relevant in 2026?

Yes, but the economics tightened. The channels that work are referral and cross-newsletter recommendation. Cold paid lists that churn inside 90 days are now a net negative on most B2B sender domains.

Is making a newsletter profitable?

It can be, but profitability is decided by retained-subscriber CAC and deliverability health, not by subscriber count. A 50,000-subscriber list with a $9 retained CAC and a degraded sender reputation is worse off than a 15,000-subscriber list with a $2 retained CAC and clean placement.

How much does SparkLoop cost?

SparkLoop charges roughly a 20% commission on paid recommendations plus 3.5% payment processing. You pay after screened subscribers confirm, which changes the true CAC math versus flat per-head pricing.

What is SparkLoop used for?

It places paid newsletter recommendations across 25+ platforms and thousands of vetted publications. In this test I used it as a buyer to acquire subscribers and benchmarked it head-to-head against Beehiiv Boosts on retained CAC.

Do paid newsletter subscribers actually retain?

Not at the rate the "confirmed" invoice implies. In this test, the Boosts cohort dropped from 22% day-30 opens to 11% by day 90, while referred subscribers held above 35%. Paid can work, but only with a screening layer and only when you price the retention curve into CAC.