As retailers face increasing competition, online recommendation quality has become a major driver of conversion and customer lifetime value. Our client, a S&P 500 listed fashion retailer, asked QuasiScience to improve their systems. We developed a suite of advanced, context-rich recommendation models, generating 3% uplift in total sales, worth several million dollars.

Consumers are expecting personalised experiences; they expect that [we] know who they are — not just that we recognise them when they are online, but wherever they are interacting with the brand.

— Mary Beth Laughton, EVP of U.S. Omnichannel Retail at Sephora

As retailers compete in an environment where customer expectations for personalisation are higher than ever, even small improvements in relevance can translate into large revenue gains. Digital channels are a key arena in which to influence purchase decisions, so recommendation systems that surface the most compelling products, improve conversion rates, and maximise cart value, are crucial to success.

Our client, a major retail brand, was using simple recommendation rules (such as ‘most bought items’) that failed to incorporate data on user behavior. This meant they could only offer static, low-context suggestions that neither reflected individual preferences nor adapted to the shopper’s intent. Our client knew that millions in potential revenue were being wasted, but lacked the modeling infrastructure to capture this opportunity.

True Personalisation

QuasiScience designed a comprehensive, multi-layer recommendation framework that captured far richer context around each customer and their product interactions. Key components included:

• Context-Rich Models: Drawing on transaction data, location data, and records of users activities to understand both long-term preferences and immediate intent.
• User and product-based recommendations: Tailored ‘you might also like’ suggestions, as well as complementary and similar item recommendations.
• Dynamic experience optimisation: so teams could test and deploy different recommendation strategies and measure their impact in real time.

We implemented these models using scalable pipelines that fit seamlessly into the client’s existing digital infrastructure, minimising the amount of training and adaptation required for staff.

Wins for both customers and the business

The project delivered significant commercial value:

• A 3% uplift in sales across various categories, totalling an additional £2m over the first year of the pilot.
• Faster, evidence-based decision-making, as merchandising, digital, and marketing teams now have clear evidence on which types of recommendations perform best.

Applications Across Industries

Recommendations are not only relevant to the world of fashion - they are an essential tool for any large-scale B2C business. And they also extend beyond the online retail environment - similar techniques can be used to offer real-time in-store personalisation, audience-level targeting, and next-generation omnichannel experiences. However you use them, one thing is clear: cutting-edge recommendation engines are crucial for competitive advantage in an increasingly data-driven retail landscape.