An average can rise because the purchase changed

## The invoice average describes a purchase

An average cost per kilogram is a compact way to describe a large set of purchases. Its compactness also makes it easy to ask the wrong question of it. If the average rises, executives may immediately discuss supplier inflation or procurement performance. Yet the company may have bought a different mix of specifications, in different quantities and on different terms. The average faithfully reflects that changed purchase without telling us which mechanism caused the movement. Before judging the buyer or negotiating with the supplier, I want the business to identify what it is comparing.

In our fictional review, average invoice cost across a selected input population moved from ₹146 to ₹159 per kilogram. The draft account described the movement as price pressure and carried a ₹2.1 crore procurement-leakage hypothesis. Procurement colleagues objected that the manufacturing requirement had changed. Finance needed a defensible basis for the next forecast. This was not a dispute that could be resolved by choosing the more persuasive department. The dataset contained a genuine movement, while the explanation attached to that movement was still incomplete.

A kilogram makes the units appear comparable, but equal weight does not mean equal economic content. Different material specifications can have different prices and different uses. Even within a familiar description, purchase terms, order size, freight treatment or another condition may have changed. I therefore resist starting with the expectation that the entire difference must be assigned to one cause. The first task is to describe the population and the accounting basis precisely enough that the comparison has a stable meaning. Only then can a method help separate possible influences.

The broad average still has value. It describes what the company actually bought on the stated basis and may be relevant to cash planning or to understanding a changing manufacturing requirement. The mistake is not calculating it. The mistake is using a descriptive buying measure as though it were automatically a judgement about the prices of unchanged inputs. Removing the average entirely would lose useful context. Keeping it beside a more controlled comparison allows management to ask two questions without pretending they are the same question expressed in different language.

## Build the comparison with the people who buy

We worked with procurement and finance on an ₹18 lakh review. The work began with item matching, comparable specifications, units and the treatment of freight and taxes. Those choices were not administrative details to be settled after the calculation. They determined what the calculation meant. A data team can match two codes and still miss a commercially significant change in what was purchased. Procurement colleagues supplied that interpretation, while finance helped define the basis on which the comparison could be used in management reporting.

For a matched fixed basket, the corresponding measure moved from ₹146 to ₹150 per kilogram. That result suggested a materially different account from treating the full rise to ₹159 as price pressure on an unchanged basket. It did not mean that the remaining nine rupees had one simple explanation. Composition changes and interactions between quantities and prices can affect the difference. Unmatched or changed specifications also require separate treatment. Naming the residual as though it were a directly observed cause would merely replace one oversimplification with another that sounded more technical.

The fixed basket had boundaries. It represented the selected comparable population and the chosen weights, not every input the business might need next year. If manufacturing requirements shift, the future purchase account may reasonably look different. A stable basket can help examine price movement while becoming less representative of actual buying. That tension is not a defect to hide; it is a reason to present the comparison's coverage and explain what it excludes. I want readers to understand both why the basket is useful and why it cannot carry every forecasting conclusion alone.

The procurement head initially wanted the fixed-basket result to settle the performance discussion. I understood the concern: the broad average had begun to sound like a verdict on the team. But the revised analysis did not establish that every procurement decision was good. It established that the first explanation was inadequate. There could still be unfavourable comparable price movements, weak terms or avoidable choices within the population. Those questions needed focused investigation. An analytical correction should protect fair judgement without becoming a blanket defence for the function whose objection prompted it.

## Do not book a hypothesis as a recovery

The ₹2.1 crore leakage hypothesis was withdrawn from the draft working account pending a better basis. That withdrawal did not recover ₹2.1 crore. No supplier paid money back because a heading changed, and no expenditure disappeared from the actual books. The ₹18 lakh investigation remained an expense. I am careful about this distinction because analytical teams can be tempted to claim the value of an unsupported number they prevented from being repeated. A corrected hypothesis may improve a decision without producing a realised financial benefit equal to the number that was removed.

Finance retained responsibility for the forecast and the accounting treatment. The revised evidence helped it decide which assumptions required further explanation, but the data team did not acquire financial authority by producing an index. Procurement retained the supplier conversation, and operations retained the explanation for changing material requirements. Our role was to make those discussions more precise. When functional responsibilities stay visible, the analysis can support challenge without encouraging each department to use a different version of the same number as a substitute for its own judgement.

The practical follow-up concerned comparable price changes, specification choices and the quantities the business expected to buy. Some questions belonged with suppliers; others belonged with the company's own planning decisions. A generic instruction to reverse inflation would have blurred those responsibilities. The better discussion asked what could actually be changed, by whom and at what cost to the product or production plan. That is where a decomposition earns its place: it directs attention towards an actionable question without claiming to have supplied the answer simply by partitioning a total.

I would change the sequence next time. Procurement should help define comparable items before an average is circulated as a management conclusion. Finance should state whether the intended use is description, budgeting, supplier negotiation or performance assessment. The same calculation may be informative for one use and misleading for another. Making that purpose explicit early costs time, but it can prevent a more expensive argument later in which departments defend or attack a number that has never had one agreed meaning.

For a board, the useful discipline is to ask what stayed constant in a comparison. If the answer is only the unit label, a claimed price movement may contain substantial changes in the thing being bought. If a fixed basket has been used, ask how much of actual purchasing it represents and what has fallen outside. Those questions do not require directors to rebuild procurement analytics. They help directors recognise whether an observed rise has been given a causal explanation it cannot yet support. The aim is a fairer decision about the business, not a more elaborate average.