The Challenge
In an increasingly competitive retail landscape, a European retail giant faced challenges in creating a seamless customer experience across its numerous online and physical store outlets. The retailer recognised a disparity in how customer interactions were being tracked — online engagements were disjointedly managed alongside in-store activities. This fragmentation hampered their ability to launch cohesive, personalised marketing initiatives. The retailer needed a robust platform that could seamlessly amalgamate data from its digital and brick-and-mortar presences, enabling precise customer segmentation and targeted marketing campaigns.
The Solution
By deploying an advanced Customer Data Platform (CDP), the retail giant aimed to bridge the gap between their online and in-store customer engagement strategies. The CDP, designed to collect, store, and analyse customer data from multiple touchpoints, provided a unified view of each customer's interactions both in the digital sphere and physical locations. Advanced machine learning algorithms processed this consolidated data to furnish powerful insights, allowing the retailer to devise personalisation strategies that catered to individual customer preferences and behaviours.
A central feature of the CDP was its integration capability. It seamlessly connected with existing systems — CRM, e-commerce platforms, point of sale terminals, and customer service dashboards — thereby creating a single repository of customer data. The CDP's analytics module enabled the retail giant to track behavioural patterns and preferences, crafting tailored messages that resonate at the right moment and on the preferred platform for each customer.
Key Results
The deployment of the Customer Data Platform brought transformative results to the retailer's marketing and customer engagement operations:
- The retailer achieved a 35% increase in customer engagement within the first six months post-deployment, directly attributable to more effective, targeted campaigns.
- Conversion rates for personalised campaigns surged by 28%, leveraging insights derived from integrated data sets.
- A reduction in marketing costs by 15% was realised, as the retailer was able to optimise resource allocation by focusing on higher yielding, well-targeted campaigns.
- The retail giant witnessed a 40% improvement in customer retention rates, bolstered by enhanced, consistent experiences across all shopping channels.
By unifying their customer data, the retail giant not only enhanced their marketing effectiveness but also spearheaded customer retention initiatives by meeting modern expectations for personalisation and convenience. This strategic pivot has reinforced their position as an industry leader adept at capitalising on the evolving digital economy.
Technical Approach
Building a CDP capable of unifying data from dozens of European markets, multiple e-commerce storefronts, and hundreds of physical stores required deliberate architectural choices at every layer of the data stack:
- Identity resolution engine: The most fundamental challenge in cross-channel data unification is the identity problem — a customer who shops online as "anne.dupont@email.fr" and pays in-store with a loyalty card number are the same person, but no single system knows that without explicit linking logic. The team built a probabilistic identity resolution engine that matched customers across channels using a combination of email address, loyalty card number, hashed mobile number, and browser fingerprint. Exact matches were resolved deterministically; near-matches were resolved probabilistically with configurable confidence thresholds, giving the marketing team control over precision versus recall in their audience builds.
- Snowflake data warehouse as the single customer data repository, chosen for its ability to handle large-scale concurrent analytical queries without contention, and its native support for semi-structured JSON data from the e-commerce event streams.
- dbt (data build tool) for the transformation layer, creating a version-controlled, tested library of data models that produced clean, business-ready customer profile and segment tables from raw event data. dbt's testing framework ensured that data quality issues upstream were caught before they corrupted downstream segments.
- Segment (Twilio) as the CDP layer for audience activation, enabling the marketing team to define segments using a self-service interface and push those audiences directly to connected channels — email platform, paid social, push notification service, and in-store clienteling app — without requiring engineering support for each campaign.
- Machine learning models developed in Python (scikit-learn, XGBoost) for propensity scoring: next-best-product recommendation, churn risk prediction, and promotional sensitivity scoring. Models were retrained weekly on fresh transaction data and scores were written back to the Snowflake customer profile table for use in segmentation.
Implementation Highlights
The implementation proceeded in three phases, each building on the capabilities established in the previous:
Phase 1 — Data foundation (months 1–3): The priority was establishing a trustworthy, unified customer record. Source system connectors were built for the primary CRM (Salesforce), the e-commerce platform (Magento), and the EPOS loyalty programme across all markets. Data governance policies were agreed with the retailer's legal and privacy teams to ensure GDPR compliance in data collection, storage, and activation — particularly important given the cross-border nature of the European customer base.
Phase 2 — Segmentation and activation (months 3–5): With a reliable customer profile in place, the marketing team began building their first cross-channel segments. Initial segments were deliberately simple — lapsed customers, high-value online-only shoppers, loyalty members with no online activity — to validate the activation pipeline end-to-end before introducing ML-driven scoring.
Phase 3 — Personalisation at scale (months 5–8): The propensity models were deployed and integrated into the segmentation workflow. The email and paid social campaigns were enriched with personalised product recommendations drawn from the next-best-product model. A/B testing infrastructure was implemented to measure campaign performance rigorously, separating the effect of personalisation from other variables such as send time and creative format.
A particularly significant challenge was navigating the retailer's fragmented marketing technology landscape, which had evolved organically across multiple acquisitions and markets. Five separate email service providers were in use at project outset. Rather than mandating a single-platform migration, the team built an abstraction layer within the CDP activation workflow that could push audiences to any connected channel, giving the marketing teams flexibility to migrate email providers on their own timeline without disrupting the personalisation programme.
Measurable Outcomes
The CDP delivered commercial impact measurable across multiple dimensions:
- 35% increase in customer engagement within the first six months — an outcome verified through A/B comparison against a hold-out group of customers who continued to receive non-personalised communications.
- 28% uplift in conversion rates for personalised campaigns versus the prior-year generic campaigns, representing a significant increase in revenue per email sent.
- 15% reduction in marketing expenditure, driven by two mechanisms: the elimination of duplicate contacts across siloed databases (reducing paid media audience costs) and the improved targeting precision that reduced spend on low-propensity audiences.
- 40% improvement in customer retention, measured over a 12-month cohort analysis. The retention improvement was most pronounced in the 18–35 age cohort, who responded most strongly to push notification and personalised social ad formats that had not previously been possible without unified data.
- The churn prediction model delivered an estimated €2.1 million in retained annual revenue in its first year of operation by identifying at-risk customers sufficiently early to intervene with targeted win-back offers before they lapsed.
Lessons Learned
Several important lessons emerged from this engagement:
- Data governance must be co-designed with legal and privacy teams, not bolted on afterwards. GDPR compliance requirements influenced several fundamental data architecture decisions — most significantly, the need to support the right to erasure at the individual customer level across all downstream systems. Designing this capability from the outset was substantially more efficient than retrofitting it post-launch.
- Self-service segmentation tools are only as good as the data models that underpin them. Early in the project, the marketing team attempted to build segments using raw event data and found the results unpredictable. The investment in dbt transformation models to create clean, well-documented, business-concept-aligned tables paid back immediately in marketing team autonomy.
- Start with simple segments before deploying ML models. The phased approach — beginning with rule-based segments and introducing ML scoring only once the activation pipeline was validated — meant that the team understood the baseline before adding model complexity. This made subsequent model performance evaluation far more meaningful.
Why This Approach Worked
The CDP succeeded because it resolved a fundamental structural problem: the retailer's customers existed as multiple disconnected personas across multiple systems, and no single team had visibility of the whole picture. A customer who purchased online three times last year and visited a store twice was simultaneously classified as a "digital customer" in the e-commerce team's reporting and an "infrequent visitor" in the store team's loyalty data. Neither view was accurate, and neither team could act on the full relationship.
By creating a single, unified customer profile and making it accessible to every channel simultaneously, the platform enabled the marketing team to engage customers as individuals rather than as audience segments — and that shift from segment-level to individual-level personalisation is precisely what drove the step-change in retention and conversion performance.
Speak with our Data Engineering team at Adyantrix to find out how we can support your next project.
Work with Adyantrix
If you are looking to tackle a similar challenge, Adyantrix has the expertise to help across the full project lifecycle. Our data engineering practice covers pipeline design, streaming, and data infrastructure. Our data analytics practice covers BI reporting and self-serve analytics platforms. Our business intelligence practice covers BI strategy, platform selection, and dashboard delivery. Our AI & machine learning practice covers ML model development, MLOps, and intelligent automation. Get in touch to discuss your requirements — no commitment required.
