SaaS Onboarding Email Teardowns

Most onboarding email advice is built from single screenshots. Someone posts a nice welcome email, everyone nods, and nobody sees the eleven emails that followed it or the fact that three of them arrived after the user had already done the thing being asked. So we did the boring version. Six accounts, six products, every send logged with a timestamp for 30 days.

What follows is the full timeline for each one, the trigger we could infer from behaviour, the activation event the sequence was clearly chasing, and the part we would change. If you want the assembled version rather than the teardown, the SaaS onboarding email sequence playbook has the structure we recommend and the onboarding email templates have the copy.

How these six sequences were captured

One fresh account per product, created between 6 and 8 January 2026, logged through 7 February 2026. Each account used a unique address on our own domain, a business email rather than a free one, a UK based signup, and a company name that suggested a 20 person software company. No credit card unless the product required one.

Behaviour was deliberately uneven. On three accounts we completed the obvious first action within an hour. On the other three we signed up and did nothing at all, which is how you find out whether a sequence branches or just runs. That difference is the whole reason the capture is useful.

Send counts below include lifecycle and marketing email but exclude password resets, receipts and security notices. That distinction matters more than it sounds: two products hit double digits only because a weekly newsletter started on day eight.

Notion: 9 emails, and the only one that mattered arrived on day zero

A PLG collaboration tool has one activation event and everybody in the category knows what it is: a second human in the workspace. Notion’s sequence clearly knows it too, and then spends most of its length on something else.

The day zero email did the work. It named three starting points rather than one, which respects the fact that nobody knows what they want a blank workspace for, and it put the invite prompt above the fold.

Then the sequence loses the thread. The invite nudge, which is the email aimed at the actual activation event, arrives on day 12. On our idle account that is eleven days after the moment it would have been useful. On the active account it still arrived, because by day 12 the sequence has switched from lifecycle to a general marketing calendar and stopped checking state.

What we would change: move the invite nudge to day two on the idle branch and suppress it entirely once a second member joins. Push the AI tour behind an event, not a date.

Vercel: 5 emails, each one tied to a deploy state

The tightest sequence in the capture, and the only one that never sent us something we had already done. Five emails in 30 days. Two of them we would call load bearing.

On the idle account, email two never arrived and the day two nudge did. On the active account, the day two nudge never arrived and the deploy confirmation did. That is branching working correctly, and it is rarer than it should be.

Developer tools have an advantage here. The product emits clean events, the team that builds the product usually owns the emails, and the audience punishes volume harder than any other segment. The lesson travels anyway: if you can only wire two events, wire the one that says the user succeeded and the one that says they stalled.

Amplitude: 12 emails chasing an event most trials never reach

The longest sequence we logged, and the clearest example of the failure mode this page exists to name. Twelve sends in 30 days, with the first at 41 minutes after signup.

The activation event is obvious for an analytics product: real data arriving from a live source. Getting there requires an engineer, a deploy and usually a meeting. That is a multi-week job for most teams, and it means a 30 day email sequence is mostly talking to people who have not yet been able to do the thing.

Almost none of it branches. The SDK install guide arrives on day eight on both accounts, including the one where data was already flowing. The strongest email in the set, the one explaining chart types, arrives on day two when there is nothing to chart.

What we would change: cut it to four sends and gate everything on the data connection event. Before data arrives, the only job is helping an engineer get the SDK in. After data arrives, the job is the first useful chart. Mixing the two produces a sequence that is wrong for everyone. This is the exact pattern we lay out in activation email sequences.

Klaviyo and Figma: the vertical case and the freemium case

These two belong together because they solve opposite problems and both solve them with volume.

Klaviyo sells into ecommerce operators, which means the buyer already knows what email marketing is and mostly wants to know whether the store integration works. Its sequence ran 10 emails, heavy on integration setup and industry benchmark content, with a clear switch at around day 10 from setup help to category education. The setup half is strong. The benchmark content is good enough that we would keep receiving it, which is a reasonable definition of a working newsletter and a bad definition of onboarding.

Figma ran 7 emails and chased file creation, then collaboration. The freemium problem is that a free user who never converts is still a valid outcome, so the sequence has no urgency in it anywhere. There is no trial clock, no expiry, no scarcity, and on balance that reads as confidence rather than weakness. The trial expiry sequence that carries a time boxed trial simply does not exist here, and nothing replaces it.

Figma’s day four email on the idle account asked one question and offered one link. Short emails are underrated in onboarding, and we counted more words in a single Amplitude case study send than in Figma’s first four emails combined.

6sense: what a sales assisted sequence does differently

We requested a demo rather than signing up, which is the only available path, and logged what came back. Eight emails in 30 days, and the shape is completely different from the five PLG sequences.

There is no product to activate, so the sequence is doing three jobs: confirm the request, get a meeting in a calendar, and keep the account warm if the meeting does not happen. Email one arrived in under a minute with a booking link. Email two came from a named SDR on day one. Emails three through eight alternated between the SDR and marketing, which is the part we would fix.

For sales assisted products the equivalent of an activation event is a booked meeting, and the sequence should suppress hard the moment it exists. Ours did not. Two more booking requests arrived after a meeting was on the calendar. If you are building this motion, the product qualified lead email plays cover the handoff rules in more detail.

The cross sequence comparison

Two numbers in that table are worth staring at. The first is the emails-before-any-action column: on the idle accounts, Amplitude sent six emails before the user had done a single thing, and every one of them assumed setup was in progress. The second is the suppression column, where three of six products never stopped.

Send count correlates with almost nothing on its own. Suppression correlates with everything. A 10 email sequence that stops at activation is a better experience than a 6 email sequence that does not.

The most common flaw is sending after activation

Four of six sequences kept describing setup after setup was complete. It is the single most repeatable defect in SaaS onboarding email and it is almost never a strategy problem. It is a plumbing problem: the event exists in the product database, nobody pushed it into the email tool, so the sequence runs on dates.

The cost is not measured in unsubscribes, which is why it survives. It shows up later, in the open rate of the expansion email and the renewal reminder that actually needed to be read. Attention spent on an irrelevant setup nudge in week two is attention unavailable in month nine, and that is the bill that lands in the churn prevention campaigns you run afterwards.

There is a real tradeoff here and the honest version is this: event driven sequences break in ways date driven sequences do not. A renamed event in a product release silently stops a send. A misfiring identify call floods a segment. We have seen a team ship a schema change on a Thursday and discover on the following Tuesday that three weeks of activation emails had gone nowhere. Date based drips fail loudly or not at all, which is genuinely why some teams keep them. Wire the events, then wire an alert on send volume per sequence per day.

Rewriting the weakest sequence we captured

Amplitude’s twelve sends are the obvious candidate. Here is the version we would run, on the same product, with the same activation event.

Four emails instead of twelve, one branch instead of none, and the entire marketing calendar suppressed during the window that decides whether the account survives. We would expect activation to move by a small amount and unsubscribes to fall by a large one, and we would be unsurprised if the activation gap turned out to be inside the noise. That result is also worth having, because it tells you the constraint lives in the product rather than the inbox.

What to fix in your own sequence this week

Pull your last 30 days of onboarding sends into a spreadsheet with the timestamp and the trigger, exactly as the tables above. Most teams have never seen their own sequence as a timeline and find at least one email nobody remembers writing.

Then do three things in order. Check the delay on your first send and get it under two minutes. Add a single suppression rule on your activation event, even a crude nightly one. Move the email that targets your activation event to the earliest point in the sequence where it makes sense, which is almost always earlier than it currently sits.

After that, decide what belongs in email at all. A meaningful share of what we received should have been an in-app message instead, shown to a user who was already looking at the relevant screen. If you want to know how your open, click and activation numbers compare before you change anything, the SaaS email benchmarks are segmented by motion and ACV, and the subject line patterns that hold up in lifecycle sends are collected separately. The rest of the cluster sits in the SaaS email marketing hub.