Dunning channel debates usually start with open rates, which is a little like choosing a vehicle by top speed. The question is never which channel is best in the abstract. It is which job needs doing at this point in a failed-payment sequence — getting attention, explaining a problem, or collecting a new card — because the three channels divide that work almost perfectly, and each is bad at the others' jobs.
Email is the explanation channel. It is where the full story fits: what failed, why it might have failed, what happens next, when access changes, and exactly what the customer needs to do. It is searchable later, which matters when a customer writes to support three weeks on claiming they were never told. Its weakness is timing — an email can sit unread for two days, and a renewal clock does not care. Email carries the record and the reasoning; it does not reliably carry urgency.
SMS is the attention channel. A text gets seen, quickly, by nearly everyone. It also gets resented when it tries to carry more than a nudge: a paragraph of explanation in a text reads as an intrusion, and a payment link with no context reads as a scam. The correct SMS is short, expected, and pointed — a failed payment exists, here is the one link that fixes it, here is the day access changes. Used at the two or three moments in a sequence where attention is the binding constraint, it earns its place. Used as a parallel copy of every email, it trains customers to ignore it.
In-app alerts are the context channel, and the most underused of the three. A customer who opens your product while their payment is failing is, at that moment, the most reachable and most motivated person in the entire sequence — they are trying to use the thing. A clear banner with a working update path converts at rates no sent message matches, because it asks for nothing: no inbox, no click-through, just the fix placed where the intent already is. Its limit is reach — it only speaks to people who show up — which is why it complements the sent channels instead of replacing them.
Sequenced well, the three stop competing. The first failure triggers the email with the full explanation and a quiet in-app banner for whenever the customer next arrives. The middle of the sequence, where silence usually means the email is buried, is where one SMS belongs. The final notice returns to email for the record, with the banner still waiting in the product. Each channel appears at the moment its strength is the thing the sequence lacks. A fuller scoring of the three, including deliverability and consent considerations, is in this comparison of dunning channels and what each one is actually for.
Two rules keep a multi-channel sequence honest. First, one source of truth: every channel points at the same card-update page and the same deadline, because a customer who gets two different dates believes neither. Second, action anywhere stops everything: a card update in the app must silence the pending SMS and email immediately, or the sequence teaches customers that your messages do not reflect reality — the fastest way to be ignored next time.
Measure the system, not the channels. Recovery by day of sequence and by decline type tells you where attention or explanation is failing; a channel leaderboard mostly tells you which message happened to be last. Build the sequence as one instrument with three voices, and judge it as one.
The sequence design, templates, and timing notes behind this are collected at Swipe Again, alongside the channel comparison linked above.
