338. The Pop-up That Decides Not to Show (and the Abandoned Cart Discount That Solves the Wrong Problem) - Taras Talimonchuk, Claspo
Every time Will Laurenson's CRO agency removes a pop-up, conversion rate goes up. The catch is that without it, the brand stops capturing emails. Taras Talimonchuk's team at Claspo has built an agent to settle that argument one visitor at a time, by deciding when a pop-up should not show at all.
Taras is Brand and Growth at Claspo, a Shopify pop-up app built around an agentic system. He has spent 16 years in digital marketing, and his team trains the agent to optimise pop-ups for revenue per session rather than email capture alone. It learns on the live traffic of agencies and Shopify merchants who act as design partners.
Revenue per session ends the argument
The idea came from a fight inside Promodo, a digital agency in Ukraine. The retention team wanted pop-ups to capture emails, and showed that each one earned around $5 on the first purchase alone. The website team wanted nothing to interrupt the main conversion goal.
Revenue per session, calculated with lifetime value included, puts both sides on one number. If a pop-up would cost more than it earns for a particular visitor, the system does not trigger it.
What the agent looks at
The agent predicts using on-site behaviour, lifecycle stage including RFM, and traffic source. Taras said it trains on a business's own real-time data, not industry benchmarks, because the best channel differs by brand. For one business it is TikTok paid ads, for another it is Google PPC.
When Meta traffic subscribes but does not buy
One retailer's data showed that Meta traffic arriving in the second half of the day was more likely to subscribe and 2.5 times less likely to buy. The agent proposed a more aggressive approach for that segment, with higher frequency and full screen pop-ups, and a lighter touch in the morning, when those visitors were going to buy anyway. "I would never find it, to be honest," Taras said.
Context still matters. A home and garden client told Taras that its Google Ads cost per click moved with the weather forecast, something no agent could have known without being told.
Holding back the incentive
The agent also tries to give away as little as possible. Its incentives include 5% off or a $10 bonus, and it will not raise the offer when it knows demand is heading down for the season. Will pointed out why that matters: a shopper who is not buying today may come back and pay full price.
Is price really why they abandoned?
Most teams answer an abandoned cart with a discount pop-up. Taras questions the assumption. "Because discount is something that works with the price problem," he said. Someone may have been interrupted, or may be using the cart as a wishlist, as he does himself.
He also cited an email platform's figure that over 80% of abandoned cart sessions are unauthorised, which means the business has no email to follow up. In that case, a pop-up is the last chance to keep a contact the business has already paid to acquire. Will added that a blanket discount goes to plenty of shoppers who did not need one.
Why Taras does not trust exit surveys
Taras does not trust "where did you first hear about us" surveys, because only certain people answer them, so the results cannot be extended to everyone. Will agreed, adding that someone buying a Jaguar might first have seen the brand 35 years ago, and that people tend to answer with the last place they saw it. Will's alternative is to interview customers about the situation of the purchase itself, including who they spoke to.
Put the hardest question third
When Taras tested quizzes, he found the question about purpose, such as whether someone is buying sneakers or comfort or status, is the hardest to answer. Placed first, it drives abandonment. His order is easy segmentation first, such as who the shopper is buying for, then preferences, then purpose. Will prefers open-ended questions where possible.
Key takeaways
- Judge pop-ups on revenue per session, including lifetime value, not on email capture alone
- Train targeting on your own traffic data, because the best channel and timing differ by brand
- Test whether a segment subscribes but does not buy, then change how aggressive the pop-up is for it
- Do not assume price is the reason for an abandoned cart, because a discount only fixes a price problem
- Treat survey answers about first touch with caution, because only some people reply
- In a quiz, ask easy segmentation questions first and the purpose question third
Taras closed on AI adoption. Fewer than 10% of the e-commerce marketers at a Bucharest conference said they used agentic systems. Listen to the full episode to hear how he thinks teams should respond.
.png)
.png)


.png)
