PMGM7023 Chap.2 What One Row Represents
What One Row Represents
A report that was arithmetically perfect
The first extended case in the course is a monitoring report recommending an immediate investigation. Every amount on the underlying file is real, every addition is correct, and the percentage change is calculated exactly. The failure sits below the arithmetic, in a question nobody asked: what does one row of this file represent?
Operational systems record contacts.
A customer opens a claim, an agent requests evidence, the customer replies, the item is collected. That is four rows carrying the same amount, each entirely genuine.
Exported without the thread that connects them, the file shows four events of that value, and the total is four times the truth in a file where nothing looks wrong.
Supplying context changes what is being calculated
Adding the missing relationship does not correct an error in the calculation. The same visible amounts, grouped one way, describe contacts; grouped the other way, they describe events.
The lecture states the lesson as a boundary rather than a technique: what gets computed follows from what is known, so widening what is known moves the result while every figure stays as it was.
This is why the case is taught first.
A group comparison, a correlation or a regression coefficient all inherit whatever unit the data arrived in, so a method applied to the wrong unit produces a precise answer to a question nobody asked.
Six lines of inquiry before one explanation
Once a number can be trusted, the temptation is to explain it, and the lecture resists that deliberately.
Six branches open first: what exactly increased and what one row means; compared with what baseline and over which denominator; whether one account is driving it; where it is located by product, site or reason; when it began and what changed operationally at that point; and what result would confirm, weaken or redirect the investigation.
Each branch needs different data, so choosing one early quietly rules the others out.
The last branch is the one that pays, because deciding what result would trigger which action removes the most common failure in applied work, which is producing a finding and then inventing the story that makes it interesting.
Rival stories are closed, not just confirmed
A genuine rise still admits several explanations, and each demands a different comparison and a different action.
A seasonal reading needs prior years aligned to the same holiday window rather than the same calendar date. A batch reading needs units sold as well as units returned, because a large batch produces a large absolute count simply for being large. An account-level reading needs peers matched on order value and product mix.
Two of three explanations being closed is a normal and reportable outcome, and a reader who cannot see the closed branches cannot tell whether they were considered.
What this chapter covers
- 01
A correct calculation over the wrong unit produces a wrong claim
- 02
Information boundary: context changes the answer without changing arithmetic
- 03
Six lines of inquiry that keep rival explanations open
- 04
Pre-committing a decision rule before the result is seen
Auditing a maintenance count before a hiring request
- 3Name the unit the file records and the unit the claim needs.
- 3Recount on the second unit and report both figures.
- 3Say which figure reaches the decision, and label the other.
Key terms
- Unit of Analysis
- Unit of analysis is the thing one row of a table describes. Every count, rate and comparison is a statement about that unit and not about any other.
- Denominator
- A denominator is the base a count is divided by to make it comparable, such as orders placed or hours worked. Without one, a larger count may only mean a larger population.
- Decision Rule
- A decision rule states in advance which result will trigger which action. It is written before the analysis so the choice is made while you are still indifferent to the outcome.
- Rival Explanation
- A rival explanation is an alternative account that fits the same observation. Each one usually requires its own comparison, so they are tested rather than ranked by plausibility.
What One Row Represents FAQ
How do I check whether a total in a business file is meaningful?
Write one sentence stating what a single row records, then ask how many rows one real-world event can produce. If a single event can generate several rows, the total counts contacts or attempts rather than events, and it has to be grouped by an event identifier before it can be compared with anything. Report both figures and label which one the decision uses.
Why write a decision rule before running the analysis?
Because after a result appears it is very easy to construct the story that makes it interesting. Stating in advance which outcome would trigger action, which would close the question and which would require more evidence fixes the interpretation at the only moment when you have no stake in it. A useful test is whether you can name a result that would make you recommend doing nothing.
Exam move
Take any table you already use at work or in another course and write its unit sentence. Then list the events that could generate more than one row and check whether the reporting handles them. Practise the six lines of inquiry on a news claim each week, writing which data each branch would need before choosing one.
Working through What One Row Represents in PMGM7023? Sia is AskSia’s AI Management tutor — ask any PMGM7023 What One Row Represents question and get a clear, step-by-step explanation grounded in how PMGM7023 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
