Two heights need a shared meaning before they can be compared

A height value appears straightforward when it arrives as a number in a familiar unit. The simplicity is useful, but it can conceal the reference that gives the number its meaning. In a combined dataset, two entries may look ready for comparison while describing height from different bases. The technology interface can place them together without noticing the problem. That is precisely why successful integration cannot be the only acceptance question. The service must retain enough context for a qualified person to determine whether the values can support the comparison that the receiving user intends to make.

Our review began with nine hundred height records. Seven hundred and eighty had the reference account required for the common view. One hundred and twenty had a different or insufficiently established basis. A delivery colleague proposed applying a constant adjustment to align the latter group. The suggestion had practical appeal: the process would be quick, the table would be complete and the client would not have to manage exceptions. It also assumed the answer to the question the practice needed to investigate. Arithmetic can produce a tidy column without establishing that its entries have become meaningfully comparable. I supported review before that apparent completion became the product.

A numerical column can hide a different question

The problem was not that transformations are inherently suspect or that all sources must begin in the same form. Bringing material into an appropriate common reference can be a legitimate professional task. The issue was the basis for choosing and accepting the method. A constant adjustment suggested for convenience was not a substitute for understanding the source, its reference and the intended use. Arun's practice had to determine what was appropriate for the questioned records. My team needed to preserve the original account and support the authorised process, rather than silently replace uncertainty with a computational operation that happened to be available.

This distinction becomes harder to defend when the software can complete the operation instantly. The visible effort of obtaining clarity may seem disproportionate to the ease of changing a value. Yet the cost of the keystroke is not the cost of knowing whether the change is justified. A manager examining progress should ask which part of the work is computational and which part depends on evidence or professional interpretation. Otherwise the team that finishes quickly by assuming meaning can appear more effective than the team that identifies the unresolved dependency. That comparison rewards a neat output at the expense of a supportable answer.

The unit was not the only required context. A value can carry a familiar label while lacking the reference description needed for the particular comparison. Conversely, a well-described source may require a qualified transformation before it can join the receiving account. These are different conditions and deserve different handling. Technology should help distinguish them rather than turn every exception into one generic error. The receiving professional needs to know whether the problem is missing evidence, a known difference requiring treatment or a value whose proposed use needs reconsideration. That distinction directs the next action and prevents repeated requests for details that cannot settle the actual issue.

Reconciliation is an accountable change

We allocated ₹ 47 lakh to source examination and supporting technology work. Qualified review reconciled 110 of the questioned records, leaving ten held. The resulting common account contained 890 supported records. These counts describe one population and its disposition, not an improvement percentage for the business as a whole. The original values and reference descriptions remained available alongside the authorised changes. That history was important because a receiving user might later need to understand why a value appeared as it did, or revisit its suitability for a purpose different from the one considered during the reconciliation.

An audit trail alone would not make an inappropriate transformation correct. Recording who pressed a button cannot replace the evidence and authority behind the decision. Equally, a professionally justified change becomes difficult to explain if the technology erases its source context. The two responsibilities therefore need to meet. The practice determines and accepts the treatment; the service carries it into the product without losing the relationship to the original material. As CTO, I am responsible for making that meeting workable. I cannot withdraw behind the claim that the system merely moved numbers when its design makes a meaningful distinction impossible for the next user to see.

The accepted common account also did not settle every question about the underlying observations. Reference reconciliation is not approval of a drainage design, an assessment of slope stability or a decision about land rights. Those uses require their own evidence and qualified owners. This limit should be clear without making the work sound worthless. Establishing comparability can be a necessary contribution to a later decision while remaining insufficient to decide it. Executive reporting often struggles with that middle position. It wants a completed outcome or a failure, when much professional work produces a specific, valuable condition on which another accountable judgement still depends.

Keep the unfinished ten in the delivery account

The client sponsor worried about delay and wanted to know whether work could proceed. A response consisting only of a warning would have been incomplete service. We needed to identify the supported set, retain the ten exceptions and explain which owners could decide the next step. A client might adjust the proposed use, obtain further evidence or postpone the affected portion. Those choices belong to the project's actual circumstances. We should not make them implicitly by giving unresolved records a plausible replacement value. A visible exception is useful only when the receiving team can understand its consequence and obtain an accountable answer about what happens next.

Holding ten records also had an administrative cost. The team had to maintain an incomplete delivery and avoid allowing a later export to reintroduce the unresolved values as though they had passed review. The commercial account needed to retain that work rather than mark the whole programme finished because most records were supported. This does not argue for endless review of every dataset. It argues for a defined acceptance decision and a proportionate way to preserve the exceptions it leaves. The alternative is not necessarily faster resolution; it may simply move the unresolved question to someone less able to reconstruct the source evidence.

My earlier client-side role made that consequence particularly clear. Receiving teams often encounter a polished deliverable long after the specialists who assembled it have moved to another assignment. They need the product to communicate its limits without relying on an informal conversation that only the original participants remember. That is one reason I favour establishing the required reference information before source acceptance. Earlier clarity cannot prevent every problem, but it can make missing context a recognised part of the service agreement rather than a surprise discovered only when a deadline makes asking questions feel unreasonable.

A board examining a technical service can use this case to ask how management distinguishes a completed transfer from a supported interpretation. The directors do not need to prescribe a vertical-reference method. They need confidence that responsibility, evidence and unresolved work are visible in the commitment being reported. Our account retains ₹ 47 lakh of expenditure, 890 supported records and ten held. It offers no broader engineering assurance. The practical standard I want from technology is simple to state and demanding to maintain: after the data has moved, the next qualified person should still be able to understand what it means and what remains unknown.