The majority of carts that get started never get paid for. For most brands the entire response to that is a single reminder email, sent hours later, saying the same generic thing to everyone. By the time it lands the intent has cooled, the customer has moved on or bought elsewhere, and the message reads as nagging rather than helpful. The revenue was sitting right there, one well-timed nudge away, and the funnel let it walk out the door.
We run cart recovery as a system that reacts in the window where it still works. Here is how it detects, times, and personalises the follow-up, and how it knows when to stop.
The system watches for more than the classic abandoned cart. It catches the high-intent browse, the repeat visit to a product page, the checkout started and dropped at shipping. Each of these is a different signal with a different best response, and treating them the same is why generic recovery underperforms. The customer who reached the payment step is a very different prospect from the one who looked twice and left, and the system responds to each accordingly.
Timing is most of the game. A recovery message fired too late is noise; fired in the warm window it converts. The system sends the first touch while intent is still hot and sequences the follow-ups across email and SMS, leaning on the channel the customer actually engages with. It references the specific product they were looking at, not a blanket you-left-something-behind, because relevance is what turns a reminder into a purchase.
A recovery system that does not know when to quit becomes a spam machine, and that costs more than the cart it was chasing. So the sequence stops the moment the customer buys, and it backs off a contact who is not responding instead of grinding through the whole series. Protecting the list is part of the job, because a customer trained to ignore you is worth less than the single sale you are chasing.
One store had a flat, single-email recovery that quietly underperformed for months. Switching to intent-based detection and a timed two-channel sequence, we found that the biggest recoverable group was not classic cart abandoners at all but repeat browsers who never made it to checkout. Catching them with a timely, specific message recovered a meaningful chunk of revenue that the old single-email flow had never even addressed.
We track recovery rate by intent type, so we know which signals actually convert and which are noise. We track revenue per recovery message, so the sequence earns its place in the inbox. And we watch opt-out rate closely, because the fastest way to ruin cart recovery is to push so hard that the customer leaves the list entirely.
The system handles the detection, the timing, the personalisation, and the send. The human sets the offer logic and the brand tone, the parts that need judgment. Recovered checkouts on autopilot are some of the cheapest revenue a store can win, because the customer already wanted to buy.
This runs on Atlas, the operating system for DTC brands. More at atlas.arthea.ai.
Occasional insights on infrastructure, conversion systems, retention architecture, and AI deployment, shared when they’re worth reading.
