What goes into a single customer view
A full SCV aggregates data from several layers:
Transactional layer — order history, average order value, purchase frequency, returns, preferred
categories.
Behavioural layer — product views, clicks, searches, time spent on product pages, add-to-cart
events that never converted, session behaviour.
Contact layer — email address, phone number, delivery address, support ticket history, NPS
responses.
Predictive layer — churn probability, predicted loyalty segment, affinity by category and brand.
Important: an SCV is not a snapshot, it is a live profile. Its value is directly proportional
to how current the data is. A profile that refreshes once every 24 hours will miss the signals that
matter most — an add to cart, a drop-off at checkout.
Identity resolution as the foundation of the SCV
The hardest technical part of an SCV is matching identifiers. One customer can interact through:
- a desktop browser (cookie)
- a mobile app (device ID / IDFA)
- an anonymous visit from another device
- an authenticated account (email / user ID)
| Identifier | Type | Stability |
|---|---|---|
| Email / user ID | Deterministic | High |
| Phone | Deterministic | High |
| Cookie | Probabilistic | Medium (gets cleared) |
| Device ID | Probabilistic | Medium (changes on reinstall) |
Without identity resolution, the purchase history from the app and the browsing history from the
desktop stay in separate profiles — and the recommendation algorithm never sees the whole picture.
The SCV and the quality of personalization
The practical effect of profile completeness: recommendations built on online behaviour alone only
capture part of the picture. Add offline purchase data and the algorithm stops recommending items
the customer already bought in store. Add support history and you can exclude the product categories
that generated returns.
For most mid-market e-commerce teams, a complete SCV that includes offline data is still a
year-plus integration project. The sensible starting point is consolidating online data and adding
offline sources as the infrastructure matures.