E-Commerce Stores Lose 96–99% of Visitors – Here's How to Recover

01 /The Concept

Most e-commerce stores watch 96–99% of their visitors leave without buying. Markopolo exists to recapture a meaningful slice of that loss – and its numbers are striking: 30–40% more visitors converted, mobile conversion up 67%, abandoned carts recovering at 8x the industry average.

The cumulative result: stores recover 35% of revenue that would otherwise be lost – $54 million recaptured across Markopolo's customer base so far.

The platform works in three layers. It starts by identifying 20% of anonymous visitors – matching them against existing customer records or enriching their data to surface contact information. It then tracks 384 behavioral signals per visitor: which products were viewed and compared, where hesitation appeared in the purchase flow, where the visitor dropped off. Those signals feed a model that infers what the visitor intended to buy and why they didn't. If a visit doesn't end in a purchase, the platform reaches out – via email, SMS, WhatsApp, or phone – with reminders and options designed to resolve whatever created the hesitation.

More than 800 e-commerce stores are already using Markopolo. The company just launched a new version, announcing it on Product Hunt.

Markopolo has raised $4M in total, including $2M closed in May.

02 /Why It Matters

Conceptually, Markopolo is attacking an old and persistent problem in e-commerce: the gap between visitor intent and actual purchase.

The scale of that gap is easy to measure. Average e-commerce conversion rates sit at 1–4%. That means 96–99% of visitors who show up at an online store with what looks like purchase intent leave without buying.

One of the main reasons: people rarely buy the first time. They browse, compare, think about it. Something eventually clicks and they return to make the purchase.

That means the store's real task isn't flooding visitors with offers. It's staying present when that click happens – or triggering the click through a well-timed, well-targeted message.

Which implies that outreach to unconverted visitors should be more personal. But "personal" doesn't mean using someone's first name in an email, or even sending them recommendations based on past purchase history. It means going back to their specific original intent and resolving whatever got in the way.

Markopolo can profile a visitor as, for example: "someone who comparison-shops prices during their lunch break, then reads reviews on their phone in the evening before buying." The platform's AI will send that person a curated set of reviews for the products they were comparing at noon – complete with direct purchase links – by the evening. A different visitor gets something different, timed differently.

Y Combinator graduate Eden ([related review](/review/kak-zastavit-rassylki-rabotat)) operates on a similar principle. Its AI analyzes visitor behavior to automatically segment unconverted visitors by the likely reason they didn't buy, then generates personalized follow-up messages designed to address that specific hesitation. If the model concluded someone was uncertain about sizing, it sends a size guide. If they balked at shipping costs, it sends alternatives or a shipping discount.

Made With Intent ([related review](/review/prodavat-nuzhno-nezhno-a-ne-grubo)) also focuses on inferring original intent from site behavior to serve the most relevant follow-on offers. It raised £1.5M in the spring of last year.

Session AI ([related review](/review/prodavat-mozhno-dazhe-neizvestno-komu), formerly ZineOne) raised $43M for anonymous visitor intent modeling on e-commerce sites.

One additional note on Markopolo: the startup is from Bangladesh and has just been accepted into the HF0 accelerator in the US. The ambition matches: Markopolo's stated mission is to have AI agents accompany every purchase within five years – from in-session guidance through post-visit follow-up. The target is recovering $100 billion in lost e-commerce revenue by 2030 and becoming a multi-billion dollar company within two years.

03 /Opportunities

Even a startup from Bangladesh can realistically aspire to become a billion-dollar company – if it points the right technology at the right market. And investors are willing to believe it, which is why a startup from outside the traditional tech hubs gets accepted into a top US accelerator.

More precisely: investors aren't betting on the specific startup. They're betting on the combination of problem, technology, and market. The specific startup matters less than the thesis, which is why smart investors typically back ten similar companies in the same space within a single cohort – if they believe the category is real.

In other words: investors bet on the theme. And the theme here is solid. There's no serious doubt that AI agents will, within a few years, be able to meaningfully analyze e-commerce visitor behavior and make offers people genuinely can't refuse.

The only open question isn't whether this happens – it's which startups will survive to lead the category.

This echoes an old piece of advice attributed to Paul Graham: "Imagine you're in the future. Look at how things work there. Then just build what's missing to get from here to there."

So: option one is to imagine that future in your own category. Option two is to take the category described in today's review and build something in it – because that future is coming regardless.