15 changes deployed across AI search, personalisation, and on-site experience. £381K annual value — every change A/B tested against revenue and margin before it went live.
Most e-commerce sites show the same products to every visitor. The search returns keyword matches ranked by algorithms nobody has touched in years. Category pages display products in a default sort order that hasn't been optimised. The homepage is a static billboard.
The result: customers who would buy if they could find the right product leave because the site didn't surface it. Search abandonment rates of 30-50% are common. Conversion rates sit at 2-3% when personalised experiences routinely deliver 5-8%.
When we overhauled search and personalisation for a major UK fashion retailer, we deployed 15 changes across AI-powered search, behavioural personalisation, social proof, and on-site experience. Every single one was A/B tested against revenue and margin before rollout. Total annual value: £381K.
Personalisation is not a single project — it is a systematic upgrade of how a site surfaces, ranks, and presents products to each visitor. We deploy it in four layers, each building on the last.
We replace keyword matching with neural search that understands intent, handles synonyms, and learns from user behaviour. Custom ranking models that optimise for conversion, not just relevance.
We deploy AI-generated product attributes for better filtering and discovery. Automated tagging that goes beyond what merchandisers can manually maintain across thousands of SKUs.
Results adapt based on browsing history, purchase patterns, and customer segments. Returning visitors see a different site from first-time visitors. High-value customers see different products from bargain hunters.
Social proof, dynamic homepages, collection-based navigation, and redesigned category pages. Every element A/B tested against revenue and margin before we roll it out.
We migrated the site's search from a traditional keyword-matching engine to an AI-powered search API with neural capabilities. The difference is fundamental: traditional search matches words; neural search understands meaning.
A customer searching for "summer dress" now sees results that include "lightweight floral midi" and "cotton sundress" — products that are semantically relevant even though they don't contain the exact search terms. The neural model learns from click patterns and purchase behaviour, continuously improving.
On top of neural search, we built custom ranking models that factor in margin, stock levels, and conversion probability alongside relevance. The search doesn't just show what matches — it shows what's most likely to result in a profitable sale.
We deployed AI-generated product attributes across the entire catalogue. The AI analyses product images and descriptions to automatically tag attributes like style, occasion, material, pattern, and fit — going far beyond what merchandisers can manually maintain across thousands of SKUs.
Customers can now filter by attributes that didn't exist in the original product data — "occasion: wedding guest," "style: oversized," "pattern: abstract" — because the AI generated them from imagery and descriptions.
The enriched data feeds back into the search ranking model, making results more precise and filtering more useful. This is where the CDP we build and personalisation intersect — better product data makes every downstream system more effective.
Search results and product recommendations now adapt based on each visitor's browsing and purchase history. A customer who consistently browses premium brands sees those brands ranked higher. A customer who buys predominantly in sale sees sale items surfaced first.
This isn't just search — it affects product listing pages, recommendation carousels, and email content. The personalisation engine builds a profile from every interaction and uses it to rank, filter, and present products in the order most likely to convert for that specific customer.
For returning visitors, the effect is immediate: the site feels like it knows them. For new visitors, session-level signals — what they've clicked, what they've searched for, what price range they're browsing — drive personalisation from the first few interactions.
Beyond search and personalisation, we redesigned the core on-site experience across multiple touchpoints.
Category pages were redesigned with improved layouts, better filtering, and collection-based navigation that groups products by curated themes rather than just taxonomy. This increased browse-to-product click rates and reduced bounce.
Social proof elements — purchase count badges ("234 bought this week"), trending indicators, and popularity rankings — were deployed across product listing pages, product detail pages, and the shopping bag. In our A/B tests, social proof delivered a 3-8% conversion rate uplift, with the strongest effect on mid-consideration products.
The homepage was rebuilt with dynamic, segment-driven content. Different customer segments see different hero banners, product carousels, and promotional content. A loyalty customer sees new arrivals in their preferred categories. A lapsed customer sees re-engagement offers. A first-time visitor sees bestsellers and brand story.
We built a wishlist feature that does double duty: it gives customers a save-for-later experience they expect, and it generates first-party data on purchase intent that feeds re-engagement campaigns. Wishlisted items trigger automated emails when they go on sale, come back in stock, or are about to sell out.
Collection point shipping was added, giving customers click-and-collect at thousands of locations. This reduced delivery costs and increased conversion for customers who prefer not to wait at home.
International shipping expansion opened new markets with localised delivery options, currency display, and duty calculation. Each market was tested with limited ranges before full rollout.
This is the critical point: every single change was A/B tested against revenue and margin before full rollout. Not just conversion rate — margin. A change that increases conversion but decreases margin per order is not an improvement; it's a trap.
We ran controlled experiments with statistical significance thresholds. Changes that didn't clear the bar were killed, regardless of how much effort went into building them. The 15 that survived represent the winners from a much larger set of hypotheses — which is exactly how it should work.
15 changes deployed across search, personalisation, social proof, and on-site experience. Every one validated through A/B testing against revenue and margin.
Personalisation and search is one capability within our ongoing work with a major UK fashion retailer. While cost-reduction work (workforce automation, warehouse, cloud) improves the denominator, personalisation improves the numerator — more revenue from the same traffic, at better margins.
Lower costs and higher revenue per visitor. That's how you move margins permanently.
Traditional search matches keywords against product titles and descriptions. AI-powered search understands intent — it knows that "summer dress" and "lightweight floral midi" are related queries, even though they share no keywords.
Neural ranking goes further: it learns from click behaviour, purchase patterns, and browse history to rank results by likelihood of conversion, not just relevance. The search gets smarter with every interaction.
Every change is A/B tested against revenue and margin before full rollout. We split traffic between the personalised experience and the control, measure conversion rate, average order value, and margin per session, and only deploy changes that show statistically significant improvement.
No gut feel. No "it looks better" justification. Data decides.
It helps significantly. A CDP unifies browsing behaviour, purchase history, email engagement, and customer attributes into a single profile that personalisation engines can act on.
Without a CDP, personalisation is limited to session-level signals. With one, you can personalise for returning visitors based on their full history — which is where the real revenue lift comes from. We build CDPs in-house as part of our engagements.
Social proof elements include purchase count badges ("234 bought this week"), trending indicators, and popularity rankings displayed on product listing pages, product detail pages, and in the shopping bag.
In our A/B tests, social proof consistently delivered a 3-8% conversion rate uplift. The effect is strongest on mid-consideration products where the customer is deciding between similar options — the social signal breaks the tie.
The search migration takes 4-6 weeks including integration, testing, and gradual traffic migration. Personalisation layers are deployed iteratively over 2-4 months, with each feature A/B tested before the next is added.
The full deployment — search, personalisation, social proof, homepage redesign — runs 4-6 months from start to completion. We front-load the highest-impact changes to deliver value early.
Every visitor who can't find what they want is a lost sale. We'll audit your search and personalisation, show you where the revenue uplift is, and deploy the changes.