Most on-site offers do the same job. They wait for a shopper to show signs of leaving, then try to change their mind.
Exit intent, basket abandonment triggers, “wait, don’t go” pop-ups. All useful, all reactive. By the time any of them fire, the shopper has already decided. You’re not persuading them any more. You’re negotiating.
There’s a better moment to act: before that decision gets made. RevLifter calls this precovery, and it’s the idea behind Tempo, the predictive engine we’ve just launched.
The trouble with reacting late
Exit intent works because it’s targeted, not because it’s early.
Reading the signals of someone mid-abandonment beats showing the same 10% off to every visitor. But it’s still a response to a shopper who’s already checked out mentally, sometimes literally.
Cart abandonment sits at around 70% across the industry, according to the Baymard Institute. That’s not a small leak to patch with a pop-up. Reacting to it after the fact only catches the shoppers who abandon slowly enough for a trigger to fire.
The ones who leave fast, or who never trip an exit signal at all, get nothing. Not because they didn’t need a nudge, but because the system was built to notice leaving, not to notice wavering.
What precovery actually means
Precovery is acting on a shopper’s likelihood to convert before there’s any sign they’re about to go.
It’s a small shift in timing with a large effect on outcome. Instead of asking “Is this person leaving?”, the question becomes “Is this person going to buy without help?” Answer that early enough, and you can act while there’s still a full range of options open, not just a last-second discount thrown at someone halfway out the door.
This matters for the industry’s bigger problem: the economics of promotions themselves. Nielsen’s research, reported by Marketing Week, found that the proportion of unprofitable price promotions was around 84%.
BCG puts the figure for retail specifically at 30 to 40% of promotions running at a loss. Most of those aren’t bad offers. It’s an offer given to people who were buying anyway.
Precovery isn’t about running more promotions or cleverer ones. It’s about running fewer, and only where they change the outcome. Spend less on discounting overall, and the same budget is moved around more efficiently.
How Tempo predicts intent
Tempo reads how a shopper is behaving in the moment: what they’re browsing, how quickly, whether they’re circling back to the same product, and how that compares with the patterns of people who went on to buy versus those who didn’t.
Three things make that prediction useful rather than just clever.
- The key is accuracy and trust. Signals are collected, predictions are made, and the results are evaluated thousands of times before a decision is made to incentivise someone. Tempo makes forecasts and then measures how correct they were. And keeps learning to get better.
- Recent behaviour is weighted more heavily than old behaviour. Someone who browses once in March tells you almost nothing about what they’ll do today, so the model pays most attention to what’s happening right now.
- Every prediction can be explained. If Tempo decides to show, or hold back, an offer, that decision traces back to the specific signals that drove it. Nothing is a guess pretending to be certainty.
Worth being clear on one distinction, since it gets muddled a lot: this is predictive. Tempo forecasts what a shopper is likely to do next. It isn’t writing anything, chatting with anyone, or making things up.
It’s making a call, backed by data, before the moment that matters has passed.
What good looks like
The underlying logic here isn’t new. It shows up whenever a business swaps a reactive default for a predictive one.
McKinsey worked with an airline that built a machine-learning system across 1,500 variables to flag which customers were at risk before problems reached the complaint stage, rather than waiting to learn through post-flight surveys.
The first application of that system delivered an 800% increase in satisfaction and a 60% reduction in churn among priority customers. Different industry, same principle: acting on a prediction beats waiting for a signal that’s already too late to be useful.
In eCommerce, the same logic applies to promotions. BCG has found personalised offers deliver roughly three times the ROI of mass promotions, and Capital One Shopping’s research suggests 62% of clothing shoppers now delay purchases until they find a discount, whether or not one was ever necessary.
Both point to the same problem: undifferentiated offers waste budget on people who didn’t need convincing and underserve those who did.
Precovery is what happens when you close that gap earlier. High-intent shoppers see nothing, because they were never going to need an incentive. The ones genuinely on the fence get something, while there’s still a decision left to influence.
The takeaway
Recovery campaigns solved a real problem: they made offers targeted instead of blanket. Precovery solves the problem after that one. It moves the moment of targeting earlier, from “they’re leaving” to “they might not convert,” which is the only point at which an offer can actually change the outcome rather than just react to it.
If your promotions strategy still waits for an exit signal before it does anything, it’s not wrong. It’s just running about a decision too late.
