Skip to content
allanninal.dev/build
Collaborate

Part 4 of 7 · Abandoned form recoverer series ~5 min read

How the one message is written

The message is short, specific, and singular, and each of those three does real work. A long one reads as marketing, a generic one reads as automated, and a second one reads as pursuit.

Key takeaways

  • Reference the specific thing: the quote, the booking, the room, the date.
  • Two hours is a reasonable delay; same-day always, next-day never.
  • One message. The second converts poorly and generates most of the complaints.
  • Reply-to a person, not a no-reply address. A good number of people just reply.
  • The opt-out is permanent and applies to every form, not just this one.

What it says

The whole message

  • Subject: Your bathroom quote — want me to pick it up?
  • You got most of the way through a quote for a bathroom in Ashford earlier and stopped at the photos step.
  • If it is easier, reply and tell me roughly what you are after and I will put something together — or carry on where you left off.
  • If you have changed your mind that is completely fine, and you can tell me not to follow up and I will not.
  • Signed by a person, with a reply-to that reaches them.

Four lines. The specificity in the first is what stops it reading as automated, and it is only possible because the basis check already established that they got far enough for there to be something specific to reference.

Reply-to a person

A meaningful proportion of the responses to this message are replies rather than clicks, and they are frequently the better outcome: somebody typing two sentences about what they want is a warmer lead than somebody returning to a form. A no-reply address throws all of that away and signals that the message was automated, which it was, and which the wording is otherwise working to avoid.

Timing

How the timing of a single recovery message is decidedA vertical chain of five steps entered by a box labelled Abandonment confirmed, with the basis checked. Step one waits two hours from the last completed step. Step two asks whether it is still not completed, re-checked now; a completion exits to Cancel because they came back. Step three asks whether it is within business hours in their local time; if not it exits to Hold to the morning, because nothing is sent at three in the morning. Step four asks whether it is still the same day or the next morning; if not it exits to Send nothing, because the moment has passed. Step five sends it, once, ever. A note says a message that has slipped past the next morning should not be sent at all.AWS ACCOUNTAbandonment confirmedbasis checkedWait two hoursfrom the last stepStill not completed?re-checked nowCancelthey came backcompletedWithin business hours?their local timeHold to the morningnever send at 3amnoStill the same day?or the next morningSend nothingthe moment has passednoSend itonce, everA message that has slipped past the next morning should not be sent at all.
Fig 1. The timing gates. The last one is unusual and correct: a nudge about something somebody was doing on Tuesday is not useful on Thursday.
  • App integration
  • Security & identity
  • Management

Why two hours

Long enough that most people who were coming back have come back, which keeps the message from being irritating. Short enough that they still remember doing it, which is what makes the specific reference land.

The expiry at the next morning is the unusual part and it is deliberate. The entire premise of the message is that it is a helpful continuation of something they were doing a couple of hours ago. Three days later it is not a continuation; it is a marketing message about a form, and the basis that justified it has largely evaporated.

Why never a second

Why abandonment recovery stops at one messageA horizontal row of five boxes. Message one: converts, modestly. Message two: converts far less. Complaints: mostly from message two. Message three: converts near zero. One, always: which is the whole difference. A note says the second message is where this category earned its reputation.WHY THE SEQUENCE STOPS AT ONEMessage oneconverts, modestlyMessage twoconverts far lessComplaintsmostly from message twoMessage threeconverts near zeroOne, alwaysthe whole differenceThe second message is where this category earned its reputation.
Fig 2. The case for a single message. The conversion curve falls fast and the complaint curve does not, which makes the second message a bad trade even on its own terms.
  • App integration
  • Machine learning
  • Security & identity
  • Management
  • Analytics

The argument for a second message is always the marginal conversion, and it is real and small. What it leaves out is that the complaints, the spam reports and the “how did you get my email” replies come overwhelmingly from the second and third, and those costs are diffuse enough not to appear next to the conversion number.

The opt-out

One link, and it does something broader than it appears: it records a withdrawal for this purpose against that person, which the basis check reads on every future abandonment across every form. Somebody who says no once is never nudged about anything again.

That is more than most implementations do, and it costs nothing because the consent record already exists. It is also the thing that makes the message honest: “you can tell me not to follow up” is only true if it applies to everything.

Next: whether any of it works.

All posts