Most digital journeys are still shaped by the highest ‑paid person’s opinion, even though only around 1 in 8 untested changes delivers a meaningful uplift. The rest are noise: well intentioned‑ “refreshes” that burn dev time, add friction, or quietly depress conversion. The alternative is continuous, compound optimisation: a rolling cadence of research driven hypotheses, AB tests, and incremental wins that stack over time. When you treat experimentation as a growth engine, every change is a bet with odds you can quantify, not a redesign you hope for the best with. An experiment that fails is not a setback; it’s‑ a future revenue loss you just avoided. Design for Today’s Benchmark: “Best in Class”, Not “Category Average” Your customers don’t compare your checkout to your closest competitor; they compare it to Klarna’s three click flows, one‑ tap ‑in-app purchases, and Netflix‑ s‑tyle “skip intro” speed. Slow hero videos, forced registration, hidden returns policies, and vague shipping details are judged against that bar, not against the legacy UX in your sector. Practical moves that reliably shift the needle include:
Making “skip the fluff” the default with prominent shortcuts to bestsellers and popular journeys.
Treating menus as commercial real estate, not corporate org charts, and categorising by how customers think, not how your PIM is structured.
Turning product pages into rich decision hubs with in scale imagery, ‑user generated‑ content, clear delivery and returns messaging, and scannable technical specs.
These are not aesthetic flourishes; they are conversion levers that have driven double-digit‑ uplifts in product discovery, subscriptions and order completion when tested properly. Symmetry, Simplicity and Trust at the Point of Money Nowhere is digital complacency more expensive than in the checkout, where roughly 7 in 10 carts are abandoned and design is often the sole cause. Hidden fees, cryptic shipping options, compulsory account creation and unrecognisable field labels all signal “risk” at exactly the moment customers are trying to say yes. Three principles should govern any high performing‑ checkout:
Symmetry: prices, discounts and offers must match from ad to landing page to basket to payment, or you trigger doubt.
Simplicity: remove every nonessential field, use familiar language, and make guest checkout the default for ‑firstt ime‑ buyers.
Trust: stack recognisable payment options with visible security cues and reassurance copy where card details are entered.
When you test these systematically, you often get “boring” changes – a date range instead of “3–5 business days”, a padlock icon next to the card field – delivering very unboring‑ gains in revenue. Personalisation, Subscriptions and the Risk of Getting There Too Fast Personalisation and subscriptions promise recurring revenue and relevance, but implemented prematurely they frequently add complexity faster than they add value. Many brands are layering quizzes, dynamic modules and subscription prompts onto journeys that don’t yet work in their generic form. The experimentation led‑ playbook looks different:
Fix the baseline journey first, then use light touch personalisation (e.g. title level paywall copy, product‑ specific USPs) to amplify what’s already working.
Validate appetite for subscription with painteddoor tests before building complex billing infrastructure; it’s common to find that only 1% of eligible customers actually want it.
Where there is clear demand, reframe “subscription” as a membership with meaningful perks, content, access and habitbuilding‑ experiences that improve retention.
In other words, don’t multiply journeys until you’ve proven the core one converts. Personalisation should be a force multiplier, not an expensive disguise for a broken funnel. Story, Language and AI: How Not to Build a Grey Car As AI makes it easier than ever to generate competent content at scale, the real risk is not that you will say the wrong thing, but that you will say nothing distinctive at all. Grey car websites – safe, neutral, endlessly similar – are the inevitable outcome when you feed generic prompts into generic models trained on generic data. Escaping that fate requires two shifts:
Treat your website as a stage, not a vending machine: foreground founder stories, values, manufacturing practices and “jobs to be done” language that explains the real role your product plays in customers’ lives.
Train AI on your own data – customer research, reviews, support transcripts, experimentation results – so it becomes a “customer champion” that reflects your specific audience, not the internet at large.
The brands that win in this environment will be those that combine ruthless, data driven optimisation with a point of view only they can own. They won’t ask AI to decide what car to build for them; they’ll use it to understand what their customers really want – and then have the courage not to ship another shade of grey.