DEV Community

Cover image for Dunning, onboarding and win-back email benchmarks, with sources
Ihor Filippov for Senlo

Posted on

Dunning, onboarding and win-back email benchmarks, with sources

Published benchmarks for dunning, onboarding and win-back emails, and how each one was measured

Baremetrics published a benchmark in June for its dunning product: 119 US B2B SaaS companies, data for May 2026, a median recovery rate of 12.7%. Recurly's billing metrics report, built on nearly 2,000 subscription sites in 2019, gives a median of 47.6%. Both get quoted as "the" recovery rate for failed payments.

They measure different things. Baremetrics divides recovered payments by the failed payments its email sequence engaged with. Recurly counts recovery over past-due invoices as a whole, which includes everything the automatic card retries fix without anyone reading an email. Retries alone account for a lot: Stunning, which has handled dunning for Stripe accounts since 2012, reports that 20 to 25% of failed payments go through when the card is retried a few times over a week.

I'm writing playbooks for four email flows that most SaaS products end up with (failed payment, activation after signup, inactive users, cancellation) and wanted a public number for each, with a named company, a sample size and a stated method. I found about twenty that qualify. They're collected below with their limits. While searching I also saw the 47.6% attributed to two companies other than Recurly, so it's worth following any figure back to where it was first published.

Flow Number Source Limit
Failed payment 12.7% median recovery Baremetrics, 119 US B2B SaaS companies, May 2026 Only payments the dunning emails went after
Failed payment 47.6% median recovery Recurly, nearly 2,000 subscription sites, 2019 All past-due invoices, retries included
Activation 68% open, 16% click on a personal email PostHog, 2024 Per-email metrics, no control group
Inactive users 14% returned, 29% of those paid WordLayouts, 1,200 users Founder-reported, no control
At cancel 753 pauses, 94 coupons Wavve, 2,590 cancel attempts Published by the cancel flow vendor
After cancel 6% back by month 2, 8% by month 6 One consumer subscription app, 2017 No win-back effort at the time

Failed payments

The Baremetrics sample sent 123,810 dunning emails in the month and recovered 10,999 charges worth $1,236,764. That works out to about 11 emails per recovered payment (my division, not a figure from the report). Their sequence starts within 24 hours of the failure, and most recoveries that follow an email come after one of the first three, within the first week. An earlier version of the report said the day-13 lockout warning was usually the most valuable message. They removed that claim when later data didn't support it.

Stunning's write-up, based on more than 50 million emails, adds detail on the emails themselves. Customers update their card either soon after the first email or after the third. Send time has little effect. Emails that look identical get ignored, so the subject and wording should change from one to the next. The update link should open a billing page without a login, because a password reset in the middle of the process loses people.

To compare your own numbers with either benchmark, compute both rates. The query below assumes two tables, and yours will have other names. failed_payments has one row per failed charge, with the time it failed and the time it was eventually paid (null if it never was). dunning_emails has one row per reminder sent, with the id of the failed charge it was about. The syntax is Postgres and the query only reads.

-- failed_payments(id, failed_at, recovered_at)
-- dunning_emails(payment_id, sent_at)
select
  count(*) filter (where recovered_at is not null)::float
    / count(*)                                         as overall_rate,
  count(*) filter (where recovered_at is not null and emailed)::float
    / nullif(count(*) filter (where emailed), 0)       as attempted_rate
from (
  select p.*,
         exists (select 1 from dunning_emails e where e.payment_id = p.id) as emailed
  from failed_payments p
  -- failures from 30 to 60 days ago, so each one had a month to recover
  where p.failed_at >= now() - interval '60 days'
    and p.failed_at <  now() - interval '30 days'
) t;
Enter fullscreen mode Exit fullscreen mode

overall_rate corresponds to Recurly's definition and attempted_rate to Baremetrics'. If the first email goes out at the moment of failure, the two are close. If it waits for the retries to finish, attempted_rate will be much lower.

Activation after signup

I found less here than for any other flow.

PostHog documented seven versions of its onboarding sequence. A personal email to engineers had a 68% open rate and a 16% click rate. Personal tips for non-technical users averaged 54% and 6%. Unsubscribes were around 0.4%, and small changes to subject lines or body copy made no noticeable difference. The post has no control group and no measured change in activation.

Heap sent different emails depending on which setup step a user hadn't completed, and reported that users who opened them converted about 10% better through the funnel. Users who open onboarding emails are more engaged to begin with, so this doesn't isolate the effect of the emails. Copyhackers reported roughly 3.5x paid conversions after rewriting Wistia's trial emails in an A/B test, with no sample size given.

So the public data shows that behavior-based onboarding emails get opened and clicked. I didn't find a published activation lift measured against a control. To get one for your product, track the share of signups that reach your key event within 14 days, by weekly cohort, before and after the change.

PostHog also stops the sequence for six weeks when a user hasn't activated after two emails.

Inactive users

These cases come from a single Freemius article of founder interviews, and all figures are self-reported.

WordLayouts emailed 1,200 users six weeks after their last login. The email offered a walkthrough video, direct access to support and a trial extension, with no discount. 14% came back, and 29% of those were paying within two months. An earlier discount email from the same company had a 41% open rate, 8% clicks, and brought back 4%.

Panto AI had also been sending discounts. They changed the email to name the setup step the user hadn't finished and offered help with it. About 18% reactivated and about 60% went back to paying. FB Group Bulk Poster sent 7,500 early free users a short email about what had changed in the product; 33% opened it and 10% of those upgraded.

The article includes a timing table without saying whose data it is. Emailing after 2 weeks of inactivity brought back 9%, of whom 22% converted. After 6 weeks it was 18% and 41%, after 3 months 11% and 18%. Multiplied through, that is about 7 paying users per 100 emails at six weeks and about 2 at either of the other intervals.

Three companies with no control groups don't settle anything, but in each of them an email with a specific next step did better than the discount email it replaced.

Cancellations

Wavve, a podcast tool, added a cancel flow with a survey and offers. In its first three months 2,590 customers started to cancel and 1,002 left the flow without cancelling. 848 accepted an offer: 637 took a two-month pause, 116 a one-month pause, and 94 a 50% coupon. 1% gave missing features or technical problems as the reason. The case study is published by Churnkey, the vendor of the flow, but it gives absolute numbers.

For emails sent after a cancellation, Hannah Parvaz describes a consumer subscription app that had no win-back effort in 2017: 6% of cancelled subscribers returned by month two and 8% by month six. The best app in its benchmark group had 35% by month six. She later A/B tested a 20% discount against a 30% one, and both converted at the same rate.

A 2023 survey of customer success teams found an average monthly churn of 6.9% at companies with win-back campaigns and 6.4% at companies without. That's a correlation in survey data and says nothing about cause, but it doesn't support the idea that running a campaign lowers churn by itself.

Paddle says its reactivation product brings back 10 to 30% of churned customers, and Churnkey says "up to 34%". Both are vendor figures for their own tools.

Numbers I left out

Several widely repeated figures didn't meet the bar. Blog posts titled "recovers 30% of failed payments" or "activation from 18% to 41%" named no customer, sample or method. Some dunning landing pages state a "73% average recovery rate" with no detail. A "62% of failed payments are recoverable" stat and a "14.7% for a four-email sequence" stat are each attributed to a known company, and I couldn't find either at the source. "A good sequence lifts activation by 20 to 30%" appears with no source at all.

Small products

With a few dozen customers there's no reason to automate these. Emailing the person whose payment failed and asking the one who cancelled for a reason takes minutes. Baremetrics did the same with every cancelling customer at a much larger size.

What the templates look like as a result

Flow Timing Stop when Measure
Failed payment 3 emails: at failure, day 3, day 7. Different subject each time, payment link without login payment succeeds recovered ÷ payments that got at least one email
Activation only to users who haven't done the key action key event fires, or 2 emails with no reaction signups reaching the key event in 14 days, by cohort
Inactive users about 6 weeks after last login, one specific next step, no discount any login returned within 30 days and still active 60 days later
Cancelled pause offered before the cancel; afterwards one email matched to the stated reason resubscribes returned in 30/60/90 days, by cancel reason

These can run on a cron job and a provider's API, in Customer.io or Loops, or in a workflow tool on top of your own provider. I'm the author of Senlo, an open-source editor and lifecycle layer that works on top of Resend, Postmark, SES or any SMTP server. It ships these four flows as workflow templates, and I'm changing two of them based on the numbers above.


Sources, checked on 16 November 2026: Baremetrics payment recovery benchmarks, Recurly Annual Subscription Billing Metrics Report, Stunning, lessons from failed payments on Stripe, PostHog, how we built our onboarding email flow, Heap, behavior-based onboarding, Copyhackers, Wistia onboarding emails, Freemius, win-back campaigns, Churnkey, Wavve case study, Baremetrics, how we reduced churn, RevenueCat, how to win back lost customers, Customer Success Collective, win-back strategies, Paddle Retain reactivations, Churnkey, winning back churned customers.

Senlo: github.com/IgorFilippov3/senlo · senlo.io

Top comments (0)