Here's something most eCommerce teams don't measure: how many of their promotions go to customers who would have bought at full price anyway?
The answer, based on research, is most of them. Studies show 83% of online shoppers will use a discount code at checkout, even when they were already planning to buy.
And according to an IMRG and RevLifter survey of major retailers, 54% of revenue comes from orders made with voucher codes. A figure that hasn't changed in five years.
That means a huge chunk of promotional spend isn't driving new sales. It's just reducing what customers were going to pay anyway.
This is the gap that intelligent offers are designed to close.
The problem with blanket promotions
Traditional eCommerce promotions haven't changed much. A site-wide banner. A pop-up with 10% off. A discount code is floating around on third-party coupon sites. Everyone sees the same thing, regardless of whether they need it.
The results look fine on the surface. The code gets used. The conversion rate holds steady. But the margin quietly erodes, and nobody asks the question that actually matters: did that offer change the customer's behaviour?
In most cases, it didn't. The customer was going to buy. The discount just lowered what they paid.
Meanwhile, 98% of website visitors leave without making a purchase. That's where the real opportunity sits. Not in discounting harder for the 2% who are already converting, but in understanding which of the 98% might convert with the right prompt at the right moment.
So what are intelligent offers?
Intelligent offers are promotions that use real-time behavioural data to decide who sees what, when, and why.
Instead of showing the same offer to every visitor, they work by reading signals from the customer's browsing session; things like what they've looked at, how long they've been on site, what's in their basket, whether they're showing signs of leaving, and whether they're a new or returning customer.
Based on those signals, the right offer is delivered at the right point in the journey. And if a customer doesn't need an offer [because they're already heading to checkout with a full basket], they don't get one. Simple.
The key difference from traditional promotions is targeting. Blanket offers treat every visitor the same. Intelligent offers recognise that a first-time visitor browsing a product page needs something completely different from a returning customer with three items in their cart.
What this looks like in practice
There are several types of intelligent offers that eCommerce teams can deploy. Here are some of the most common:
- Exit campaigns trigger when a visitor shows signs of leaving, moving their cursor towards the close button, switching tabs, or sitting idle for too long. Instead of losing that customer entirely, a well-timed offer gives them a reason to stay and convert.
- Invalid code campaigns catch customers who try to apply a third-party discount code that doesn't work. Instead of letting them leave frustrated, a replacement offer keeps them in the funnel.
- Limited-time incentives create urgency on specific products, particularly useful for shifting overstocked inventory without running a site-wide sale that damages brand perception.
- Stretch-and-save offers encourage customers to increase their basket value. A shopper browsing with a £30 cart might see an offer for 10% off when they spend £50. It lifts AOV while keeping the discount targeted to customers who need a reason to add more.
- Cross-sell offers appear when a customer has bought one item but skipped a complementary item. (Think: someone buys running leggings but not the matching top.) A gentle nudge at the right moment can increase items per order without feeling pushy.
Each of these responds to a specific behaviour. None of them requires giving every visitor the same discount. That's the point.
The question most teams forget to ask
If you run a promotion and it gets used, was it successful?
Most teams would say yes. But redemption doesn't tell you whether the offer changed the outcome. A customer who was going to buy at full price used a code they found, that's not a successful promotion. That's the margin you gave away for free.
Intelligent offers solve this with control groups. A percentage of qualifying visitors don't see the offer. By comparing the conversion rates of the two groups (those who saw it and those who didn't), you can measure the actual incremental impact.
This shifts the question from "did this offer get used?" to "did this offer change behaviour?" It's a small reframe, but it changes everything about how you evaluate promotional performance.
What the results look like
Brands using this approach are seeing measurable gains. A few examples:
- Radley: 15% conversion rate uplift and 8% reduction in promotion costs. Proving that smarter offers can drive more conversions while spending less
- Splits59: 133% increase in conversion rate from targeted cross-sell and exit campaigns
- Shop4Runners: 108% CVR improvement after closing gaps in their abandonment strategy with targeted offers
- Flow Kayaks: 15% conversion rate increase confirmed through control group testing over a 61-day trial
These results don't come from offering bigger discounts. They come from offering the right discount to the right person at the right time. And importantly, from measuring whether it actually worked.
Where to start
If you're running promotions today, you don't need to overhaul everything overnight. But there are a few questions worth asking:
- What percentage of your promotional spend goes to customers who would have bought anyway?
- Do you measure incrementality, or just redemption?
- Are you showing the same offer to new visitors and loyal customers?
- Could you reduce your discount rate while still maintaining (or improving) conversions by targeting more carefully?
If the answer to any of those makes you uncomfortable, intelligent offers are probably worth exploring.
Promotions are one of the biggest levers eCommerce teams can pull. The challenge isn't whether to use them. It's making sure they work as hard as they should.
