A common customer view is often described as if completeness were its principal virtue. The enterprise wants fewer disconnected records and a clearer account of the relationship across stores and online activity. Those are reasonable aims. They become less useful when the programme treats every unresolved link as a failure to be eliminated, regardless of the evidence. Two records can look similar without representing the same person, and a confirmed identity link can still leave questions about how the combined details should be used. A credible design keeps those questions distinguishable rather than ask one matching decision to settle them all.
In a Varneli exercise, we evaluated 120,000 candidate links. The agreed rules supported 108,000 initially, while 12,000 remained ambiguous. Further review confirmed 4,000 of the ambiguous group and left 8,000 separate. The final counts therefore described 112,000 linked candidates and 8,000 unresolved candidates within the examined set. They were not necessarily a count of unique people, and they were not a newly authorised marketing audience. I make those distinctions explicit because a project measure can change meaning quickly once a larger number becomes attractive to another part of the business.
Keep uncertainty visible in the link
A matching rule expresses a judgement about evidence. It should be possible to explain why the rule is considered sufficient for its intended use and what happens when the evidence conflicts. Similar contact details may be useful indicators, but their meaning depends on context. I do not want the technology team to resolve ambiguity by choosing whichever merge makes the common view appear most complete. The relevant owners need to agree how uncertain cases are examined and what should remain separate. A visible unresolved state can be more dependable than a confident link whose basis nobody can defend.
Our ₹95 lakh review programme preserved the ambiguous group rather than lower the requirement for linkage. That choice had a cost and left some service inconvenience unresolved. I did not pretend otherwise. It also avoided turning a known uncertainty into an apparently settled identity. The 4,000 links confirmed through further examination had a recorded basis. The remaining 8,000 could receive later evidence through an owned process. A programme should be able to make that kind of bounded progress without being judged solely by the proportion of entries it has merged by a particular date.
The scope of the exercise matters as much as its result. We examined candidate links produced through a defined process. A candidate may involve records already connected in other ways, and several candidates may concern one person. It would therefore be misleading to describe the count automatically as the size of the customer base. I ask programme reports to retain the unit being counted. This is a simple discipline, but it prevents a technical reconciliation measure from becoming an unsupported commercial statement when the figures are reused in planning or a presentation to directors.
Separate identity from the intended use
Even when a link is well supported, it does not tell the enterprise everything about what may happen next. A store colleague helping with an order, a service team investigating an enquiry and a marketing team proposing a communication have different purposes. The company needs to examine the authority and conditions for the intended use through its appropriate process. The matching programme should not silently create that authority by placing the records together. I do not treat a successful link as a legal-compliance certificate, and I would not ask a technology score to replace the assessment required for a particular use.
This distinction should influence the application design. If a combined view makes every detail equally available for every purpose, the interface may encourage assumptions the organisation has not examined. The relevant teams need to specify what a user needs, what the user is permitted to do and where a question should be escalated. Technology can then implement an arrangement that reflects those decisions. The work is more useful when it begins before the system has been built around a broad idea of completeness, because narrowing an established expectation can be harder than designing a clear purpose from the start.
Service colleagues also need an explanation they can use when records remain separate. They should not have to improvise a claim that the customer is wrong because the system has not linked an entry. The interface can show the available evidence and an appropriate route for review without presenting uncertainty as an accusation. I want the programme to consider that conversation as part of its operating design. A technically correct unresolved status is still incomplete service if the next colleague has no way to explain what can happen or who is responsible for the next step.
The commercial temptation is understandable. A more complete-looking view can promise easier service and better coordination. I do not dismiss those benefits. I ask the sponsors to distinguish the benefit of a dependable link from the benefit they hope to obtain through a later use. Each may require different evidence and decisions. If a programme claims both at once, its investment case can become difficult to evaluate: a result in record matching is used to justify a customer outcome that has not yet been examined. Separating the propositions makes the next experiment more precise rather than less ambitious.
Design for correction after the project
Customer records do not remain fixed after a reconciliation exercise. New entries arrive, details change and earlier assumptions may be challenged. A useful system needs a route for correction that preserves the basis of the decision and makes the consequence understandable. I want to know who owns that route, what support the person has and how a mistaken linkage can be examined without relying on the memory of the original project team. Continuing responsibility is part of the cost of the common view. It should not appear only after the implementation budget has been closed.
Measures should reflect this continuing work. The number of linked candidates is one observation, but it does not by itself establish service quality or appropriate use. The programme can examine whether unresolved cases reach the right owner, whether corrections are understood and whether the intended service tasks become more dependable. Those observations should keep their scope and period. I would rather have a smaller set of well-explained measures than a broad claim of customer understanding built from an aggregate number whose relationship to actual people is unclear.
The practical ambition is a record that helps colleagues without pretending to know more than the enterprise has established. Leaving 8,000 candidates separate was not the end of the work, but it was an honest account of where the work stood. The CIO can make that uncertainty manageable by providing a clear technical state, an owned review process and a distinction between identity and use. A common view becomes valuable when people can understand and correct it, not merely when the last visible gap has been filled.