The answers up front: welcome runs 3 to 6 (built as 5 or 6 with an engagement-gated tail), abandoned checkout and cart run 3, browse abandonment runs 1 to 2, post-purchase runs 4 to 6 across its phases, winback runs 3, sunset runs exactly 2, replenishment runs 3, dunning runs 3 on top of retries, back-in-stock runs 1 plus conditional follow-ups, and cross-sell runs 3. One principle generates every number: sequence length follows decision length, and the tail of any sequence sends only to people still engaging. The rest of this post is the evidence, flow by flow, with each build linked in full.
Welcome: 3 to 6, engagement-gated
The data band is settled: 3 to 6 emails, with complex and education-heavy products earning the long end, multiple welcome emails adding up to 51% more revenue than one, and engagement decaying predictably across positions, which is normal rather than alarming. Our builds run 5 to 6 with the last two gated to engaged non-buyers, so the uninterested experience four and the warm experience six: the supplement build shows the pattern, and the vertical variants flex the tempo (F&B compresses to ten days, home stretches to three weeks) without leaving the band.
Abandoned checkout and cart: 3, then stop
Three touches inside 48 to 72 hours is the published standard, and the benchmark gap is the argument for building the full sequence: average flows convert 3.33% of recipients while the top decile converts 7.69%. Email four and beyond buys unsubscribes, not orders: someone who ignored three reminders decided. The checkout build and cart build carry the skeleton; the high-AOV exception stretches to 4 over a week because the decision itself is longer, per the considered variant, which is the principle showing itself: the count followed the decision.
Browse abandonment: 1 to 2
The earliest-intent audience gets the lightest touch: one service-toned email for fast categories, a second only for considered ones, frequency-capped per profile, per the browse build. Browse behavior recurs; the flow must not.
Post-purchase: 4 to 6, in phases
Counted across its marketing arc (the transactional stack is its own system and does not count against the sequence): the start-right email, the check-in, the review ask, and the bridge, with vertical additions like the fit-check or the purge conversation. The count is generated by the product's verdict clock: F&B compresses because taste verdicts land in days; considered purchases add wait-phase touches because the delivery arc is months.
Winback: 3, then the pipeline moves on
The published structure and our builds agree: soft re-engagement, moderate incentive, capped final offer, with diminishing returns and rising list risk beyond three. Every vertical variant (supplement, apparel, beauty, F&B) keeps the count and changes the argument; the considered build swaps the drip for occasion-triggered touches entirely.
Sunset: exactly 2
Klaviyo's two-part model: the re-permission and the goodbye, then suppression. A third sunset email is a contradiction in terms, per the build.
Replenishment: 3, clocked to consumption
The reminder at 70 to 80% of the cycle, the at-empty streak email, the past-empty recovery nudge, then exit, per the supplement build and its beauty and CPG extension. The engagement numbers that justify the flow (50 to 60% opens, 10 to 15% conversion) are earned by timing, and a fourth nag spends them.
Dunning: 3 emails on top of the retries
Retries do the silent majority (well-configured retries alone recover 45 to 70% of failed payments); the three emails front-load into the first ten days because the recovery curve is steepest in the first three days and decays sharply after two weeks, per the build.
Back-in-stock: 1, plus conditional follow-ups
The notification is the flow; the 24-hour still-available nudge sends only while stock holds, and the consolation only on repeat sell-out, per the waitlist build. At 7.9% email conversion, the moment needs no sequence, just speed and fairness.
Cross-sell: 3, gated on belief
The pairing, the economics, the objection pass, then stop, entered only after the first product earned belief: the shared skeleton across stacks, routines, outfits, pantries, and rooms, with the entry gate and tempo changing by vertical and the count holding.
The principle underneath all of it
Sequence length follows decision length: fast decisions get compressed sequences (F&B recovery in 48 hours), long decisions get patient ones (high-AOV recovery over a week, home welcome over three), and every tail is engagement-gated so the flow flexes per recipient instead of averaging everyone into annoyance. When a count question is not answered above, derive it the same way: how long does this decision actually take, who is still engaging at each step, and what does the next send add that the last one did not.
Frequently asked questions
How many emails should a welcome series have?
3 to 6, built as 5 or 6 with the tail gated to engaged non-buyers. The band is data-backed and the gating is what lets one build serve both the uninterested and the warm.
How many abandoned cart emails is too many?
Four. Three touches inside 72 hours is the standard; past that, unsubscribes outrun orders, with the single exception of high-AOV considered purchases, which earn a fourth touch across a longer week.
Do more emails mean more revenue?
Up to the decision's length, yes; past it, no. The multi-email premium is real in every flow's data, and every dataset also shows the returns collapsing once the sequence outlives the decision it serves.
What is an engagement-gated tail?
The last one or two emails of a sequence send only to recipients still opening or clicking, so persistent interest earns persistence and everyone else exits early. It is the mechanism that resolves almost every length debate.
Where do these counts come from?
Each flow's full build on this site, with the published benchmark or study behind every number linked in place. This page is the index; the builds are the evidence.