Survey, Administrative and Flow Data

Three records may describe different events

A growers' group compares a household survey, delivery invoices and warehouse weights. The survey asks what people obtained last week, the invoices record transactions, and the warehouse tracks physical movement. Their totals differ. That does not immediately prove that someone reported falsely: the records may concern different people, dates and stages.

Read a dataset's definition before reading its largest number. What event was counted: an order, a sale, receipt, preparation or consumption? Who could enter the record, and who could be absent? A useful connection between sources begins by identifying those differences rather than choosing the figure that best supports a preferred story.

Surveys depend on questions and coverage

A survey obtains answers or observations from selected people or units. Its question wording, recall period, sampling and response pattern shape what the result describes. Food purchased by a household is different from food eaten by each member. A response about difficulty obtaining food is different from a measured nutrient intake.

Retain the questionnaire or measurement definition and explain how the sample represents the intended population. A large response count alone does not establish representation. Dates matter too: a seven-day recall and a calendar-month invoice file should not be presented as observations of one identical period without an explicit justified adjustment.

Operating records have their own purpose

Administrative records arise while an organisation performs a task: placing orders, issuing invoices, recording stock or administering a service. They can provide detailed repeated observations, but their original purpose may not match the research question. A billed order can be cancelled, partially delivered or received after its invoice date.

Check identifiers, status updates and revisions before counting unique events. An order line and its later correction should not automatically become two meals or two shipments. Small producers or unregistered channels may be outside an organisation's coverage. 'No entry' can mean no event, an event outside the system or an unrecorded event; the interpretation needs evidence.

A flow account connects stocks and movements

For a stated product and boundary, opening stock plus inputs must be reconciled with outputs and closing stock, allowing for relevant transformations. Keep a common mass basis. Fresh produce with changing water content cannot silently be treated as the same dry-matter account, and processed products need appropriate conversion information.

Consider an invented warehouse-period record for one vegetable on a common fresh-mass basis, stipulating no processing, water exchange or other physical mass change. Opening stock is 20 kg and received stock 80 kg. Dispatches are 70 kg, recorded discard 5 kg and measured closing stock 20 kg. The recorded input side is 100 kg; the recorded output-and-closing side is 95 kg. Five kilograms remain unexplained by these records.

Investigate an unresolved balance

The 5 kg difference is not automatically an additional measured discard. It might reflect an omitted movement, a stock-count error, a misplaced date or another recording problem. The stated assumptions might also be wrong. A balance provides a consistency question; it does not identify which explanation is correct.

Check the weighing method, units, time cut-off and stock identities. Preserve the initial record and document verified corrections. If a later verified closing count were 25 kg rather than 20 kg, the arithmetic would reconcile. That hypothetical possibility illustrates one explanation; it is not evidence that this warehouse actually made that error. A closed account also cannot independently prove that every entry is accurate.

Report what was observed and adjusted

A combined report should distinguish directly recorded values, estimates and assumptions. Describe any conversion, imputation or exclusion that changes the result. Keep the original reporting period and corrected version accessible so later comparisons can use the same rule.

Choose the next source according to the unresolved question. Customer receipt records can clarify a delivery status; repeat stock measurement can investigate a warehouse discrepancy; an appropriately designed survey can examine access beyond registered sales. Shared conservation and uncertainty principles support these checks, while specialised nutrition, logistics and statistical methods remain in neighbouring topics.

Worked examples

Check your understanding

1. Does a household purchase survey directly measure what each household member ate?

   Show answer and reasoning

   No.

   Acquisition and individual consumption are different quantities.

2. Does an arithmetically closed warehouse account prove every constituent record is accurate?

   Show answer and reasoning

   No.

   Offsetting errors or incorrect assumptions can still produce a closed total.

3. If verified closing stock were 25 kg in the invented case, what would be the residual?

   Show answer and reasoning

   Zero kilograms.

   100−(70+5+25)=0; this conditional calculation is not a finding that the original count was wrong.

Key ideas

• Identify the event, unit, coverage and period behind each dataset.
• Surveys, administrative records and material accounts answer complementary questions.
• A residual signals an unresolved account, not a proven cause.
• Preserve original records and distinguish measured corrections from assumptions.