A graduate is not automatically a lost customer

Our repeat-customer chart showed ninety bookings from a prior group of four hundred customers. The other 310 appeared as a large space in the story of a relationship we had failed to maintain. I had helped create that interpretation because repeat purchase is an attractive way to describe marketing performance. Student storage made the limitation difficult to ignore. A person might finish a course, leave the town or no longer need the service. That context mattered. It did not mean we could remove those people from the original population and present a more flattering number as though nothing had changed.

Keep the original population in view

The 2024 review began with a straightforward overall result:ninety out of four hundred, or 22.5%, had booked again. We authorised₹27 lakh of appropriately bounded customer research to understand the non-repeat group. Of the 310 people who did not return, 180 had completed their course or left the service area, seventy were known local continuing customers who did not repeat, and sixty remained unclassified. The ninety repeaters were also known local continuing customers. Those distinctions enriched the account, but the starting four hundred remained the population whose overall repeat behaviour we had reported.

A colleague proposed leading with ninety out of 160, or 56.25%, among the known local continuing subgroup. That figure answered a potentially useful question. It did not answer the same question as ninety out of four hundred. The subgroup excluded 240 people, including sixty whose circumstances were still unknown. Presenting its rate as the overall result would make the business appear to have improved without a single additional customer returning. I retained the original 22.5%headline and showed the subgroup separately, with the basis of its classification and the people outside it visible.

The distinction is not a preference for an unflattering number. A broad overall figure can be insufficient for a business decision, and a subgroup can reveal something worth investigating. The requirement is to keep the question attached to the answer. If management wants to understand repeat behaviour among people still able to use the local service, it needs an account of how that group was identified. If it wants to evaluate the original customer cohort, it needs the original denominator. Both views can be useful when their relationship is explained rather than silently changed.

Use context to ask a better question

The 180 known exits did not establish that those customers were satisfied. A person can finish a course and also have had a poor service experience. Equally, the seventy local continuing non-repeaters did not all represent dissatisfaction or a competitor win. Their need may have changed, or they may have found another practical arrangement. Classification helps select the next question;it does not necessarily supply the cause of behaviour. I wanted the research team to retain that humility because an appealing explanation can become a new assumption as quickly as the old retention chart did.

For marketing, the context changes the work worth considering. A completed academic need might make a future acquisition programme more relevant than repeated persuasion aimed at the same person. A local continuing customer might provide useful service feedback, subject to appropriate contact and consent. An unknown circumstance requires proportionate inquiry or an honest acknowledgement that the evidence remains incomplete. These are different decisions. Treating all 310 non-repeaters as one failed relationship could lead the business to spend on the wrong activity while claiming to be disciplined about retention.

That does not make customer acquisition automatically preferable to retention. New customers also cost money to reach, and the quality of the service still matters to people whose need is temporary. A company should understand the economics of its actual use pattern rather than borrow the ideal of an indefinitely recurring customer from another business. The storage period, the academic cycle and the geography all influence what a realistic relationship looks like. My responsibility is to help the enterprise examine those influences, not to select whichever marketing measure makes the current function appear most successful.

We also had to consider how the context was learned. A useful business question does not grant unlimited permission to track a person's location, education or plans. The review used appropriately authorised evidence within its defined purpose. We did not treat students' personal movement as a resource marketing could monitor indefinitely. Qualified privacy and service owners retained the relevant decisions. Some uncertainty remained because the business did not have an appropriate or sufficient answer. That is a legitimate condition of customer research, not a defect that should always be solved by collecting more personal material.

Do not turn unknown circumstances into a convenient exit

The sixty unclassified cases mattered particularly to the subgroup story. It would have been easy to assume that they resembled the known exits and remove them from the apparent opportunity. It would have been equally easy to count them all as dissatisfied customers and use the result to argue for a larger retention budget. Neither inference was established. I kept the unknown group visible because it defined a limit on the conclusion. The business could decide whether further proportionate research was worthwhile, but it could not make the uncertainty disappear by assigning a convenient label.

The revised account made planning more complicated. We no longer had one number that seemed to describe a single marketing problem. We had an overall observation, several known circumstances and a remaining unknown group. I considered that a better basis for discussion even though it gave my team more explaining to do. A simple measure is valuable when it preserves the relevant decision. When it combines different situations into an apparently uniform failure, it can make activity look accountable while obscuring why the activity should help.

A board could use this example to ask how management defines the customer relationship underlying a growth plan. What repeat behaviour is economically plausible for this service? Which departures are expected, and how are they learned about? Are subgroup figures being presented alongside the original population? What remains unknown? Directors do not need to run the research or decide an individual contact strategy. They can examine whether management's interpretation supports the proposed investment and whether the account changes its definitions after the results become inconvenient.

I would agree the treatment of expected exits before the next cohort is formed. The team should know which questions it intends to answer, what evidence it can appropriately obtain and how incomplete classification will be reported. That preparation would make the later discussion less dependent on defending or replacing a disappointing headline. A graduate may no longer need local storage, and understanding that can improve the business. The improvement lies in the decision the context enables. It does not lie in making a customer vanish from the population whose behaviour we originally set out to understand.