The decisions you made once and nobody has checked since

Edition 052 · 8 October 2026

The expensive decisions in a business are the ones nobody remembers making: made once under conditions that have since changed, with nobody asked to check that they still hold. Five speakers at Meet Magento UK 2026 covered four of them: payments, the growth model, how you sell, and how readable you are to an AI agent. The advice is the same for all four: give each decision an owner and a date, and write down the one question that tells you if it still holds. This week I want to talk through those four, with the question I’d ask of each.

Payments: the set-up you configured at go-live

Payments is the cheapest to check, and it has had news this year. On 2 July, Visa said agent-initiated payments are live at independent merchants, including Frasers, with banks across Europe including the UK, using passkeys to meet strong customer authentication. On 30 September, Mastercard announced a probability score that tells card issuers how likely it is that a payment was started by an AI agent rather than a person. Adobe’s session said you can’t really check out through an agent in this region yet. Either way, it’s worth finding out who last looked at the authentication, fraud and settlement rules behind your checkout, because they were set up for a person at a keyboard when you went live, and they’ve probably not been touched since. They sit with three different owners (your payment provider, your fraud tool, the card issuer), so it never becomes anyone’s job to reopen them.

Janak Kika and John Baxendale from Adobe ran the payments session, and John’s own story sums up the problem. He moved to Barcelona fourteen years ago, and coming back showed him how out of date his picture of the UK was:

“In my head, the UK is still as it was 14 years ago when I left.” – John Baxendale

Open banking shows the drift: it now reports over 18 million active users. If you priced pay-by-bank at go-live, or ruled it out, the numbers behind that call have moved.

John’s list of where a retailer can recover margin works as a checklist: currency conversion on cross-border sales, interchange set up for a market that regulation and Brexit have since reshaped, settlement timing, reconciliation, and authorisation rates.

A payment set-up rarely fails visibly. The cost turns up a few basis points at a time on each order, and finance sees a fees line while the checkout looks like it works. Get someone to price last quarter’s transactions from scratch, split by currency, card type and decline reason, meaning what you’d be quoted for each line today, and put the result next to what you pay now. You’ll need that split from your payment provider, and if your pricing is blended and they can’t give it to you, that’s your first answer.

A bad answer is “we can’t split that”, or a figure that doesn’t reconcile to your statements. Then get your provider, whoever owns fraud, and finance into one conversation.

Know which levers are yours: exemptions and fraud rules are mostly your provider’s and acquirer’s, while interchange and settlement timing sit in the contract, so check for lock-in before you ask for changes. Price first, then change deliberately.

Watch Kika and Baxendale’s session in full

Growth model: sell everything, everywhere

Matthew Lawson is Chief Digital Officer at Ribble Cycles. He was on the AI panel I wrote up in edition 050, and his own talk had a different lesson. When he joined Ribble, he told the room, “the job was to grow everything”: every channel, every marketplace, every territory, every product. Sales rose. Two or three years in, the team saw they were selling more and making less.

“We have done the impossible,” he said. “We’ve made a cycling charity.”

He put the choice as going bust or changing. Behind the headline, a set of products were making money, and they were Ribble’s own bikes. So they pivoted to selling those direct to customers and walked away from 75 per cent of their revenue.

The brief was reasonable when it was set. It went unchecked because the headline number kept rising, and revenue is the easiest line to watch. I’d put margin by channel and by product in front of whoever owns growth every quarter, and ask the version of the question that stings. Revenue-only reporting is how a growth brief goes unchallenged for years. Ribble makes its own bikes, so for you it may mean one channel or one range, not three quarters of the business.

Watch Lawson’s session in full

How you sell: the product and what surrounds it

Last week’s edition featured Evgenii Garde of Hilti and his talk on growing the accounts you already have. This is the part I held back: what Hilti decided it was selling.

His starting problem was price. He put one of Hilti’s tools at three to four hundred, and said that money buys five to seven cheaper versions on a marketplace such as Amazon. Hilti changed the product instead of arguing about the price. Customers pay a fixed monthly fee for their whole tool fleet, and the fee covers repairs, a backup tool while theirs is away, dashboards, and theft cover that Hilti prices itself, much as an insurer would. It took ten to fifteen years to build. Hilti stopped competing on the item.

What I take from it is that when a customer sets you against something cheaper, the answer is often in the whole bundle before it’s in the price: how the product is presented and categorised, its data and imagery, and the service around it. Garde’s point is that the proposition is a big part of the product, so protect your margin before you reach for the price. The test I’d run: what do we bundle that a marketplace seller can’t, and are we charging for it?

Watch Garde’s session in full

AI readiness: the decision youve already made

You’ve made this one without noticing. David Williams of DHW Digital is a digital transformation consultant with twenty-five years in ecommerce, and his talk was about how AI agents find you, in three stages.

The first is being found: how AI models represent your brand, which comes down to schema, feeds, content, reviews and citations. The second is being understood, where AI helps shoppers compare and choose on your own site. The third is action: agents handling quotes, reorders and service work behind the login, with human approval and an audit trail. Product data sits under all three.

The order is what stuck with me. I read it as a comment on where effort goes: a chatbot is easier to buy than a clean product feed. His example was a client that published a page of FAQs without the schema markup, so AI systems weren’t reading them.

If you’ve never looked, you’ve still made this decision, and your platform and your product data are the answer. I’d ask your platform team four things:

Is product schema on the product page template, and is it consistent when your theme and your extensions each add their own?
Is the feed complete, and who signs off that it is? “It should be” is a bad answer.
Are FAQs marked up so a machine can read them?
Is anything in your bot protection turning agents away?

Then try it yourself. Ask an assistant ten questions a customer in your category would ask, and see if it names you and gets you right. That’s a smoke test, not a measurement, but it tells you which way to look.

Watch Williams’s session in full

Four decisions to reopen this quarter

For each one, write down what the decision was, who made it, and when anyone last checked. If the last two come back blank, it has no owner. Then ask the question.

Payments: if we priced last quarter’s transactions from scratch today, would we choose this set-up?
Growth model: which part of our revenue would we turn down if it arrived as a new customer today?
How we sell: what do we bundle that a cheaper version can’t match, and are we charging for it?
AI readiness: if an assistant were asked to recommend something in our category this week, would it find us and get us right?

The growth model is the only one of the four that changes the size of the business, and Ribble only found its answer by looking at margin by product, so start there, then payments, which is the cheapest to check. If you run several brands or platforms, give each decision one owner across all of them, or it gets checked differently in each and owned by nobody. Send the questions out this week: payments to finance, growth to whoever owns the revenue number, how you sell to whoever owns the proposition, and AI readiness to your platform team, with a review date against each.

Every other talk from Meet Magento UK 2026 is on our YouTube channel, all in one playlist. Lawson’s and Garde’s sessions are also part of The Official Meet Magento UK 2026 eBook, our written-up collection of this year’s sessions, produced by JH and Corefinity.