The Challenge
Careem’s evolution from a single ride-hailing app into a GCC super-app was a product success story — and a data engineering challenge. Each new vertical (rides, food delivery, grocery, Careem Pay) had been built rapidly, often with its own user identity model, event schema, and analytics instrumentation.
The result: a customer who used Careem Rides three times a week and ordered Careem Food daily was invisible to the personalisation layer as a single person. CRM campaigns ran on vertical-specific segments. Churn models only saw partial behaviour. The growth team was making retention decisions without being able to see the full customer.
The brief was to fix this — without disrupting the product velocity that had made Careem one of the GCC’s most-used apps, and within a data architecture that had to operate across UAE, KSA, Pakistan, and Egypt.
Our Approach
We began with a Crawl phase structured around two questions: what does the data actually look like across verticals today, and what would a unified identity unlock commercially? The answer to the second question was unambiguous — cross-vertical customers had 3.4× the lifetime value of single-vertical users, but there was no activation mechanism to grow that cohort.
The Walk phase ran over five months across three workstreams:
-
Customer identity resolution — We designed a probabilistic identity graph that stitched together first-party identifiers (phone number, device ID, app user ID) across verticals. The graph handled the complexity of shared devices, multiple phone numbers, and partial sign-in states — common patterns for GCC super-app users.
-
CDP implementation — We implemented a Customer Data Platform to serve as the single source of truth for unified customer profiles, real-time event streaming from all five verticals, and segment activation to downstream channels (push, in-app, CRM). We built the schema to be extensible as new verticals were added.
-
CX orchestration model — With unified profiles in place, we designed a cross-vertical lifecycle model: onboarding sequences that introduced new verticals at the right moment, cross-sell triggers based on behavioural signals (e.g., a regular Rides user who had never opened Food), and win-back flows that recognised churned single-vertical users before they went dark. Arabic and English variants were built in parallel.
The Outcomes
Six months after CDP go-live:
- 41% improvement in cross-vertical customer retention — measured as the percentage of single-vertical users who adopted a second vertical within 90 days of a cross-sell trigger.
- 3.1× increase in real-time personalisation volume — campaigns that previously ran on weekly batch exports now ran on streaming event triggers, reducing time-to-activation from 48 hours to under 4 minutes.
- Unified customer profiles for 94% of active users — up from a baseline where fewer than 30% of users had any cross-vertical identity linkage.
- Segment export latency reduced from 6 hours to under 10 minutes, enabling the growth team to act on same-day behavioural signals.
- CRM revenue contribution increased 28% as lifecycle campaigns shifted from vertical-specific promotions to cross-product moments of value.
The Careem data and growth teams now operate from a single customer view. Cross-vertical program planning happens in a unified dashboard rather than five separate analytics instances.
“The identity resolution work unlocked visibility we hadn’t had since the early days of the single app. Emerge Digital navigated the complexity of our multi-market, multi-vertical architecture without missing a deadline.”
— Director of Growth, Careem