A telecom business can count subscriptions accurately and still describe its reach inaccurately. The problem begins when the object counted disappears from the sentence. A registered subscription becomes a customer, then a person, then evidence of population reach. The arithmetic may remain unchanged through every step, while the conclusion becomes progressively stronger. I think the data leader’s responsibility includes resisting that movement, especially when the largest available number is also the most attractive one to communicate.
My fictional 2023 example contains 7.8 million registered mobile-subscription records. An internal ninety-day activity condition identified 6.1 million with qualifying activity and 1.7 million outside the condition. Those figures describe the review population and its defined subsets. They do not establish how many unique people used the service, how many people held multiple subscriptions or how many unobserved users shared access. No subtraction between the figures can answer those questions without further evidence about the relationships involved.
Name the object before choosing the headline
The first question is therefore not which count is most impressive, but what each record represents. A subscription can be the correct unit for a commercial or operational enquiry. It need not be apologised for because it is not a person. The mistake is to borrow the emotional or strategic meaning of a people count while retaining only subscription-level evidence. That changes what a reader believes without changing the number they see, which is why ordinary arithmetic checks may fail to expose it.
In the case, Farah Mervanti questioned a proposed statement about people reached. The communications sponsor wanted one headline that could travel easily between presentations. I understood the desire for simplicity: a qualified account demands more attention from both writer and reader. But simplification has a limit when it changes the entity. We retained the registered population and created an explicit activity account. The result was less portable as an unqualified slogan and more useful as evidence for different, clearly stated questions.
This does not require a long methodological appendix beside every figure. A concise label can carry the object, period and condition if those have been decided before the chart is designed. The fuller account should then explain the source, exclusions and limitations. The problem becomes harder when a headline is approved first and the evidence is asked to justify it afterwards. People become invested in wording, and a necessary correction can sound like an attempt to diminish the business rather than describe it accurately.
Treat activity as a condition rather than a new identity
An activity rule identifies records meeting specified conditions within a window. It does not, by itself, transform subscriptions into individuals or resolve the reasons other records did not qualify. In our invented review, the ninety-day condition was chosen for an internal purpose. I would not present it as an official statistical definition merely because it resembles a familiar indicator. If an external reporting requirement is relevant, its actual definition and applicability need separate examination by the responsible owners.
The 1.7 million records outside the condition also require care. It would be tempting to call them lost customers, because that makes the difference sound like a clear management issue. But the account does not establish that interpretation. The records may need further enquiry into their status and use. An absence of qualifying activity within one window is the observed condition; a claim about loss adds a commercial conclusion. The latter might eventually be supported, but it is not supplied automatically by subtraction.
Window changes create another source of confusion. A longer observation period can include more activity without proving that the service gained users during the interval. A changed event definition can move records between groups without indicating a behavioural change. I would preserve the definition used for each published internal comparison and explain material revisions. Otherwise a series that looks continuous can conceal changes in its counting rules, making apparent growth or decline partly a product of the account itself.
This discipline is not an argument for freezing every definition forever. A measure can become unsuitable as the service changes, and a better question may require a different population. The obligation is to make the change intelligible. Readers should know whether they are seeing a movement in observed behaviour, a change in the population or a revised method. Sometimes more than one is involved. A clear account can acknowledge that mixture without pretending to isolate an effect the evidence cannot separate.
Design the account around the claim it can support
Our S$3.8 million data work maintained the reviewed population and its activity condition. It did not create new customers, establish a national penetration rate or demonstrate a commercial improvement. I want that distinction visible when the work is described as an outcome. A reporting correction can matter because it changes a decision’s starting point. Its value does not need to be inflated into a service or revenue result that no corresponding evidence has measured.
For a board reader, I would make the principal claim easy to test. What is being counted? Which period and inclusion condition apply? What additional relationship would be needed to turn the count into the proposed conclusion? Those questions help identify the real uncertainty. They also respect the reader’s time: a definition should clarify the consequential boundary rather than invite an unlimited tour through every source field. The data leader must exercise judgement about which distinction actually changes the decision under discussion.
A commercial colleague may reasonably ask for a people estimate rather than a subscription count. That is a new evidence problem, not an instruction to relabel the existing one. It may require different sources, permissions and assumptions, with their own limitations. I would scope that enquiry openly and let the relevant owners assess whether it is appropriate. The institution should be able to decide that the extra information is useful, unnecessary or unavailable without disguising the current count as an answer it already possesses.
I would also make the correction durable in the documents people reuse. If the explanatory note remains only in one analyst’s file while the chart circulates alone, the original interpretation can return. The record should contain a short usable definition and a route to the fuller account. Reviewers should be able to see why the wording matters, rather than treat it as an optional caveat that can be removed when space is tight or the audience is senior.
My position is that precision about the counting object is part of leadership communication, not a technical obstacle to it. A subscription account can say something substantial about the service without pretending to count people. Keeping the activity condition attached allows a useful comparison while protecting against unsupported claims about reach or loss. The stronger executive narrative is the one that explains what the evidence means and why that meaning matters, even when the resulting headline is smaller than the one the room initially wanted.