Picture a retailer whose stores closed in March 2020. A customer who had visited weekly, whose face the cashier knew, who had accumulated loyalty points, now lands on the website. The site does not recognise her. She is treated as a new visitor. Her purchase history is invisible. No recommendations appear for the categories she regularly browsed. She tries once more, gets the same blank slate, and on the third attempt switches to a competitor. This is not a hypothetical. It is the structural break that many Turkish retailers encountered as footfall collapsed and digital traffic surged overnight. The problem was not a lack of technology investment; it was that customer memory had been locked inside a single channel.
A Customer Data Platform (CDP) is the architectural response to exactly this problem. The core idea is straightforward: collect behavioural data from every channel a customer touches — website, mobile app, point-of-sale, call centre, email — and consolidate it into a single unified profile that every touchpoint can read in near real time. In practice, building that consolidation requires serious data engineering and organisational coordination. It is also worth being precise about how CDP differs from CRM. A CRM is a record system that sales and marketing teams manage manually; a CDP is a data layer that assembles behavioural signals automatically and exposes them to downstream systems via open interfaces. Conflating the two often leads companies to fund an expensive CRM upgrade and call it a CDP project.
The word ‘omnichannel’ has been repeated so often in recent years that its meaning has blurred. The pandemic sharpened it again. Being omnichannel does not mean being present on multiple channels; it means enabling a customer to begin an experience on one channel and continue it seamlessly on another. The prerequisite for that continuity is that the customer’s identity and behavioural history are accessible independently of the channel. In Turkey, the number of retailers operating at this maturity level remains limited. Even large chains frequently run separate systems per channel: store POS data sits in one platform, e-commerce data in another, and the two datasets never merge. A customer who is loyal in-store is a stranger online.
The rapid growth of e-commerce in Turkey during the pandemic made this structural gap visible in revenue terms. Retailers needed to recognise their most loyal physical-store customers the moment those customers arrived on a digital channel. Some attempted a partial fix by matching card numbers or phone numbers at login — a workable but incomplete approach, because customers do not always log in and do not always use the same device. This is where CDP’s identity resolution layer matters. Identity resolution is the process of linking the same individual across different devices, sessions, and channels using deterministic matching (confirmed identifiers like email) and probabilistic matching (behavioural patterns and device signals). The technical complexity is real. And a candid caveat is necessary: identity resolution is never one hundred percent accurate. The goal is to reduce the unrecognised customer rate to a level where the business impact is manageable, not to eliminate it entirely.
For a mid-sized Turkish retailer or service company, the investment question is not abstract. Enterprise CDP platform licences and integration costs, denominated in foreign currency against a backdrop of exchange rate pressure and persistent inflation, can reach levels that strain technology budgets significantly. Recommendations that ignore this reality do not survive contact with a finance committee. However, CDP should be approached as an architectural decision rather than a product purchase. Several companies have achieved a meaningful unified-profile capability by building synchronisation between their e-commerce platform, CRM, and marketing automation tool, without procuring a dedicated CDP platform. This path is slower and less automated, but it is a credible starting point. The strategic question that guides resource allocation is: between which two channels does the handoff failure cause the most measurable customer loss? Starting there concentrates effort where it produces the clearest return.
A dimension that CDP and omnichannel discussions frequently underweight is organisational readiness. In most retail organisations, store operations, e-commerce, and marketing run on separate budgets and separate KPIs. When a customer starts a journey in a physical store and completes it online, whose revenue line does the sale appear on? Until that question has a clear answer, technology investments that facilitate channel switching face internal resistance, because one team’s gain is another team’s reported loss. The pandemic accelerated some of these structural changes: a number of companies merged channel budgets, others created cross-functional customer experience roles. But this transformation is far from complete, and the organisational friction remains a more common barrier to omnichannel execution than the technology itself.
A practical decision framework for leaders looks like this: can you currently measure where a customer begins their journey and where they abandon it? Do you know your channel-switching recognition rate — that is, what proportion of customers who move from one channel to another are correctly identified on arrival? If those two questions cannot be answered with data, the CDP conversation is premature. The productive sequence is to build the measurement infrastructure first, identify which channel handoff is causing the most attrition, and then select a data architecture approach proportionate to that specific problem. Carrying customer memory across channels is as much a strategy and organisation question as a technology question. Buying the technology before resolving the strategy is the most reliable path to an expensive system that no one uses.
This article was originally written in Turkish by Gökhan MERCANOĞLU on July 6, 2020 and has been automatically translated into English and other languages using machine translation.