eDelivery Expo 2026 | Speaker Interview
The delivery experience starts at the search bar
An interview with Mark O'Brien, VP of Business Expansion at Luigi's Box
Fulfilment efficiency usually gets treated as a logistics problem you solve in the warehouse, with the carrier, or on the last mile. Mark O'Brien thinks a good part of it is decided earlier, the moment a customer types something into the search bar. Before his session at eDelivery Expo, we asked him why findability belongs on the fulfilment agenda, what search data says about demand, and where AI actually changes the maths.
1) You say the delivery experience doesn't begin in the warehouse. So where does it begin?
Mark: It begins the second a customer types something into your search bar. Most conversations about fulfilment start at the warehouse, the carrier, the last mile. All of that matters, but there's a part of the journey that rarely makes the agenda, and it sits right at the front: search and discovery.
A customer searches and gets nothing back, so there's no order to fulfil at all, and you already paid to bring that person in. Or they search, the wrong product shows up, they buy it, it doesn't match what they wanted, and it comes back as a return. A return costs you two to three times the original fulfillment. The most expensive parcel you'll ship this year is the one that comes back.
2) You call these hidden fulfillment costs. Why hidden?
Mark: Because they don't land in the part of the P&L where fulfilment costs are supposed to live. A wrong product that comes back gets logged as a return. Rising "I couldn't find it" messages get logged as customer service volume.
The costs are real and they sit in the same P&L, they just get blamed on the wrong line. The team that owns search almost never gets the bill. So the spend stays invisible and the cause stays unfixed, while the warehouse team gets measured to the penny. That gap is the actual problem.
3) Why does search end up neglected in so many retail businesses?
Mark: Search gets switched on once and then forgotten. It sits with IT, or with whoever built the site, not with the people who answer for margin. So one end of the journey gets optimised constantly and the other end, the bit that decides whether there's an order at all, goes untouched for years.
Nobody made a bad call here. The issue is that nobody owns search as a commercial lever in the first place.
4) Your talk treats the search bar as a demand signal. What do you mean?
Mark: Every query a customer types is a statement of intent. At scale, those queries are one of the cleanest reads on demand a retailer has, and almost nobody uses it.
No result searches tell you which products customers expect you to stock and you don't, which is a direct brief for the buying team. A spike in searches for a category usually shows up days before it shows up in sales data, so it works as an early warning for planning and procurement. And the words customers actually use, instead of the words your buyers picked when they labelled the catalogue, are a valuable input for the SEO/GEO and paid teams.
Most retailers watch sessions, ROAS, email open rates, category page performance. Very few look at no result terms, or search conversion by query cluster, or search volume against current stock. The data is sitting in the logs. It carries a lot of signal and barely anyone touches it.
5) Can you give a concrete example of a search signal turning into an operational decision?
Mark: A spike in searches for "garden furniture" is a cue for the buying team to raise the stock order before the season fully lands. No search results for something specific, say "size 14 wide-fit boots", tells you there's a catalogue gap and a range worth extending. Rising searches for "eco packaging" can feed straight into a procurement brief for next quarter. A sharp drop in searches after a promotion ends is a reason to revise the forecast and move warehouse resource somewhere else.
None of that needs new data collection. It's connecting what customers are asking for in real time to what the operation does next.
6) Where does AI actually change the maths, rather than just adding novelty?
Mark: The work which used to need a team of merchandisers keeping synonym lists in spreadsheets and hand-tuning boost rules can now run mostly on its own.
A good search and discovery system learns from clicks, carts, and purchases, and improves ranking without anyone stepping in. It works out that "kids waterproof jacket" and "children's rain coat" mean the same thing, even when your catalogue only knows one of them. It copes with typos and finds relevant results when the exact keyword is missing.
The part that matters to anyone running a budget: the same team can handle a bigger catalogue, across more markets, through more seasonal peaks, without adding headcount at the same rate. Fewer wrong orders means less pressure on returns processing, on customer service, on reverse logistics. The savings work their way back through the whole operation.
7) Let's talk numbers. What's the starting point for most retailers, and what changes when search is done well?
Mark: Most people underestimate where the market already sits, so start there. We pulled 2026 e-commerce search benchmarks from 4,000 online stores. Even among shops that run advanced search, 6.3 percent of searches still come back empty. When a search fails, 32 percent of shoppers leave on the spot, and a lot of them go buy the same thing from a competitor. Compare that with the shoppers who do use search: they convert 21 percent more often than people who only browse, and the ones who use autocomplete convert at three times the rate. Search is already deciding a real slice of revenue before anyone touches it.
8) The scale problem must be getting worse as catalogues and markets grow.
Mark: It is. A large retailer might be running anywhere from 50,000 to several hundred thousand SKUs. And seasonal ranges change all the time, often across markets that each want different rankings. Asking a merchandising team to manage relevance, ranking, promotions, and seasonal priorities across all of that by hand, in real time, isn't realistic.
Automation is what lets you do more without growing the team at the same rate. Bestsellers and high-margin items rise automatically on the business rules you set. Language coverage grows from the query data instead of a hand-kept synonym list. And you can A/B test ranking and layout at the search layer with statistical confidence and no developer time.
9) If a retailer in the room wants to act on this, where should they start? It sounds like it could be a big project.
Mark: It isn't a big project to start. Ask your team three things.
What is our no search result rate, and what are the top 20 no result search terms this month? Who is reviewing our search data as a demand signal, and how often? And what share of our merchandising decisions still needs a person to do it by hand, and what would it mean to halve that?
If nobody can answer the first one in under a minute, that tells you where the attention has been going, and where the easy wins are. Fulfilment efficiency is usually framed as a logistics problem you solve at the end of the journey. A good chunk of it is a discovery problem you can solve at the start. The delivery experience is only ever as good as the discovery experience that came before it.
About Mark O'Brien
Mark O'Brien is VP of Business Expansion at Luigi's Box, where he works with e-commerce retailers across Europe on search and product discovery. He speaks regularly on the commercial and operational side of findability.
About Luigi's Box
Luigi's Box provides AI-powered search and product discovery for e-commerce, including autocomplete, product recommendations, analytics, and guided selling. It works with retailers across Europe, from mid-market to enterprise, in fashion, home, beauty, and B2B, with most clients seeing measurable impact within two to six weeks. More at luigisbox.com.
Mark O'Brien is speaking at eDelivery Expo, NEC Birmingham, 1 July 2026.