Most abandoned carts are not decisions. Someone got distracted, the page was slow, or they meant to come back later. A reminder catches some of them. The question is what the follow-up teaches.
Here is the sequence we shipped in StoreFlux, and the part I am not sure about.
The sequence
Three emails, measured from the cart going quiet. The defaults are one hour, 24 hours, and 72 hours. The sweep checks each cart before it sends, and it stops the moment the shopper completes an order, so nobody gets a reminder about a purchase they already made. Each stage is recorded, so re-running the sweep does not send the same stage twice.
Where the discount goes
The discount can only be issued on the third email. It also needs two things to be true: the store has the include-coupon setting on, and the Coupons module is enabled for that tenant. If either is off, the third email goes out without a code.
We put it last on purpose. A discount in the first email tells people the price drops if they wait a day. That is a lesson we did not want to teach.
The other side
I can see the argument against us. For some shoppers, a cart from three days ago is a second chance, and a discount is what gets them to finish. Holding the discount until the end may mean the offer arrives after the shopper has already moved on, while the people who would have bought anyway get nothing extra.
I do not think there is a clean answer. That is why this is a discussion post, not a feature announcement.
Questions
Does a recovery discount train shoppers to abandon carts on purpose? If it does, where would you draw the line: no discount at all, a discount only for returning customers, or a smaller one at the end? What has actually worked or failed on your own store?
Tell me in the comments. I would rather hear a strong disagreement than a polite nod.
Top comments (4)
Putting the discount last makes sense, and I'd add one cap: one recovery coupon per customer or email address in a set window, say 90 days. Without it, a shopper who learns the pattern can abandon every cart, wait 72 hours and collect a code each time, which is the training effect you're worried about.
The other thing that would settle the argument is a small holdout that never gets the third email, so you can see how many of those carts would have closed anyway. Does the sweep record enough to run that split?
Thanks Gautam, both points are useful, and I agree with them.
On the cap: we don't have one yet. Each abandoned cart gets its own single-use code, and nothing limits how many codes one customer can receive across carts. A per-customer window like your 90 days is the right fix, and I'd like to add it.
On the holdout: the sweep doesn't keep one. Every cart gets the same sequence. What we do record is which emails each cart received and whether it converted within an attribution window. That lets you compare, but it's observational, not a clean test. A real holdout needs a random split at the cart level, and we haven't built that yet.
Thanks for the honest answer. For the holdout, it doesn't need to be big to be useful: sending 1 in 10 carts no discount email at all, chosen at random, would show how many of those carts convert anyway. That one number tells you how much of today's recovery the discount is actually buying. Would your store's volume give you a readable result in a month or two?
This is a better test than the one I had in mind, Gautam. A random 1 in 10 with no third email is simple to run and easy to explain, and it answers the question directly.
Honest answer on volume: I don't have the numbers to say whether a month or two would be readable. That depends on how many carts a store gets. What we'd need to build first is random assignment at cart creation, recorded so the split can be analyzed. The sweep doesn't do that yet, so right now we can only compare carts, not run a clean test.