Product recommendation engine lifts email conversion by 25%

QuantSpark built a recommendation engine for a menswear retailer that personalised repeat-buyer emails against each customer's own purchase history. The result: a 25% higher conversion rate on those emails, open rates close to double the retailer's non-personalised benchmark, and a fully automated pipeline now reaching more than 100,000 customers.

The problem the engine solved was hiding in plain sight in the retailer's own data. A large share of customers repurchased similar products year after year, and customer research confirmed the instinct: more than 80% of repeat buyers wanted an item similar to their first purchase. But the retailer's marketing could not serve that instinct efficiently. Frequent changes to the product range meant a reactivated repeat buyer typically needed five or more clicks to find the item they actually wanted, and the emails prompting that reactivation were only partially personalised, rarely reflecting what the customer had bought before. Demand that should have converted easily was instead lost to friction, one click at a time.

QuantSpark's approach started with the data, not the email template. The team analysed transaction history to confirm which customers were genuinely habitual repeat buyers and what drove their choices, size, colour, fit and style, and built a recommendation algorithm that scored and suggested products to each customer individually, prioritising those statistically most likely to buy again. That model was powered by combining transaction logs, stock levels and CRM records, so every recommendation reflected real customer history and only surfaced products actually available to buy. The engine was integrated directly into the retailer's existing CRM rather than bolted on as a parallel system, and QuantSpark also supported the redesign of the marketing content itself: each repeat-buyer email was rebuilt so that at least 50% of its creative related directly to the customer's prior purchase. The whole pipeline, from data ingestion through to email assembly, was then automated end to end.

That automation is visible in the workflow itself. Transaction logs, stock levels and CRM records are pulled together automatically for each customer; the recommendation engine scores available products against that customer's own purchase and style history; the highest-scoring recommendations are woven into the email creative; the personalised email is dispatched at scale with no manual build step; and the reduced friction for repeat buyers shows up as higher open rates and, ultimately, a 25% conversion lift. No stage of that chain requires a marketer to hand-pick products or assemble a campaign for a segment: the system decides, assembles and sends.

Underneath the workflow sit four categories of system working together: the retailer's existing CRM platform, which supplies purchase history and now also receives the engine's output; the recommendation engine itself, which ranks products against individual preference profiles; the email marketing platform used to deliver the campaigns; and the data integration pipeline that keeps transaction, stock and CRM data synchronised so recommendations stay current. None of these is a bespoke, proprietary product invented for this case; each is the kind of system most retailers of this scale already run, which is part of why the approach is repeatable elsewhere in the sector.

The value delivered is best read as three linked figures rather than one. Reach came first: personalised recommendations now go out to more than 100,000 customers, entirely through automated pipelines. Engagement followed: open rates on these personalised emails run close to double those of non-personalised sends. And conversion followed engagement: a 25% higher conversion rate, attributed directly to the reduced friction repeat buyers now experience. Turning that chain into a pound figure would require the retailer's baseline conversion rate, average order value and campaign frequency, none of which is disclosed here, so no revenue number is claimed. The reliable method is to apply the 25% uplift to the retailer's own repeat-buyer email revenue base and campaign volume once those figures are known, net of the cost of building and maintaining the engine and its CRM integration, rather than to manufacture a return the available figures do not support.