Nobody knows why the shop is failing
The claim
Delivery is the most reported online purchase problem in the EU — and probably not the most important one. For an independent shop, delivery is more likely a conversion constraint than a logistics cost.
What this episode argues
In 2025, 63% of EU online buyers reported no problem at all. Among those who did, “delivery speed slower than expected” was named by 20% — the single most frequent issue. Next came a hard-to-use website (12%), a wrong or damaged delivery (10%), difficulty finding guarantee information (8%), and costs higher than indicated (5%).
Four of those five are about uncertainty, not speed. The customer does not know what will happen. And a shop that treats delivery as a courier setting is adjusting the most adjustable part of the purchase while treating the least adjustable part as if it were adjustable.
We test this against Eurostat purchase-problem data and Belgian retail structure, then hand the claim to a shopkeeper who is invited to break it.
The counterfactual test
If delivery were the binding constraint, then changing how delivery is presented at checkout — and nothing else — would move conversion measurably. If shops that changed delivery presentation saw conversion move in proportion, and shops with clean, fast, well-communicated delivery still were not converting, then the constraint is elsewhere and we have said something misleading.
We have no data either way, and we say so on air.
What we would need to know
- Delivery cost as presented at checkout versus as charged, for one shop, over three months.
- The share of orders where the stated delivery date was missed.
- What customers actually said when they abandoned checkout.
- Order value and margin before and after a delivery-presentation change.
Most shops already have the first and second of those. Almost nobody has the third.
One thing we got wrong
When we started this project we assumed the useful thing to measure would be some kind of authority score — how well a retailer is positioned around a concept. We built it, and we found you could raise that number by repeating keywords and lower it by writing more informative copy. That measures nothing a shopkeeper cares about. That is why we are not selling it, and why the first field note is about delivery rather than about AI visibility scores.
- Eurostat, E-commerce statistics for individuals (data extracted February 2026) — eurostat.ec.europa.eu
- KBC Economics, Retail vacancy rate in Belgium at record high, 20 October 2026 — kbc.com
- Landmark Global, Top 10 Essential Facts About Belgium Ecommerce: 2026 Edition — landmarkglobal.com
Transcript
Jason: Welcome to The Counterfactual. A podcast about retail and retail technology, from the people who want to be wrong in public.
Jason: I'm the host of this show. There's no research institution behind this, no university, no corporate sponsor. There is one small research association in Belgium, and it has no clients, and it does not sell software. I'll come back to why that matters, because it's the whole point of the show.
Jason: Each episode works the same way. We start with a claim. One sentence, specific enough to be arguable. Then we look at what supports it, what contradicts it, and what we could not find. Then we hand the claim to somebody who is invited to break it. Then we ask the counterfactual question: what would have had to be true for the claim to be false? And finally, we say one thing we got wrong.
Jason: Episode one. My name is Jason. My guest today is someone who actually runs a shop, and we're going to spend most of this conversation on the question of delivery, because it is the question everyone in retail is being told about, and almost nobody has actually tested.
Jason: Before we get to that, here's the claim. In 2025, sixty-three percent of European online buyers reported no problem at all with buying online. Of the ones who did report a problem, the most common one by a distance was delivery speed being slower than expected. Twenty percent named it. Next came a website that was hard to use, at twelve percent. Then a wrong or damaged delivery, at ten. Then difficulty finding information about guarantees and legal rights, at eight. And costs higher than indicated, at five.
Jason: Those numbers come from Eurostat, from the ICT households survey, data extracted in February 2026, covering 2025. That's a solid source. I'm not going to pretend otherwise.
Jason: Now here's the claim, and this is where we push back on our own data. We think delivery speed is the most reported problem, and it is probably not the most important one. And for an independent shop, delivery is more likely to be a conversion constraint than a logistics cost.
Jason: Let me be precise about what we are and are not claiming, because this is the part where research usually goes wrong. We are not claiming delivery does not matter. We are claiming something subtler. If you read the complaints list again — delivery too slow, site hard to use, wrong delivery, guarantee information hard to find, costs higher than indicated — four of those five are about uncertainty, not speed. The customer does not know what will happen. The promise is unclear, or unclear late, or unclear in a way they discover at checkout.
Jason: And here's the thing about a shop that treats delivery as a shipping setting. You're adjusting the most adjustable part of the purchase, and treating the least adjustable part as if it were adjustable.
[stage direction — the guest introduces themselves and their shop; we ask: what does delivery actually mean in your business — a courier contract, a promise on your site, or something else?]
[stage direction — the counterfactual test: if delivery were the binding constraint, changing how delivery is presented at checkout, and nothing else, would move conversion measurably. Has that ever happened in your experience?]
[stage direction — what we would need to know: delivery cost as shown at checkout versus as charged over three months; the share of orders where the stated date was missed; what customers actually said when they abandoned; order value and margin before and after a change.]
[stage direction — one thing we got wrong: we assumed the useful thing to measure would be an authority score. We built it, and found you could raise the number by repeating keywords and lower it by writing more informative copy. That measures nothing a shopkeeper cares about.]
Jason: If you run a shop and you have a decision you're stuck on — one decision, one period, one unit, one shop — we will analyse it for you, for free, in your language, with the method and the limits shown, and no sales pitch afterwards. We don't take customer data. The link is in the show notes, and it's also on retailresearch.eu.
Jason: The claim was: delivery is more likely a conversion constraint than a logistics cost, and speed is not the thing that matters most. What would have had to be true for that to be wrong? If shops that changed delivery presentation saw conversion move in proportion, and shops with clean, fast, well-communicated delivery still weren't converting — then the constraint is elsewhere and we said something misleading.
Jason: Next time: we take on the AI recommendation you should distrust, and I bring the receipts on our own failure. The Counterfactual is produced by Retail AI Research vzw, a research association in formation in Belgium. Sources, transcripts and methods are published openly at retailresearch.eu. If you run a shop and want us to look at one real problem, the link is in the show notes. Thanks for listening.
Run a shop? Bring us one problem
One decision, one period, one unit. We will send back a written analysis in your language, free, with the method and the limits shown — and no sales pitch afterwards. We do not take customer data.
Open the diagnostic form ↗