# Reconciliation practice with an audit trail

LearnWithZavi | Synthetic practice

Use two small fictional exports to identify an extra duplicate row, a missing reference and an amount mismatch. Your result is a checked discrepancy log, not an instruction to change a real ledger. This practice extends the existing reconciliation lesson; it does not require AI, a paid tool or an external account.

## Read the rules first

All records, source notes and amounts in this pack are invented. FC means fictional credits, not a currency. The scenario date is 3 November 2026, a teaching date rather than a publication date. Both exports cover the same fictional account, day and unit. Positive values increase the practice total; negative values reduce it. These conventions are for this exercise only.

Keep both raw files unchanged. Copy them into a separate practice workbook or inspect the printed tables. A row ID identifies an export row; reference identifies the underlying practice item. Use exact reference matching within account and unit. Do not match different references merely because amounts are equal.

Flag a reference that appears more than once before any one-to-one match. For this pack, a clean match has one row in each export, equal reference, date, account, unit and signed amount. Keep duplicate groups outside the clean-match count until reviewed. No fuzzy matching, rounding tolerance or automatic date shifting is needed.

## What you will hand in

1. Row and reference counts for both files, plus raw signed totals. 2. A clean-match list and a discrepancy log with row IDs. 3. A signed bridge explaining B minus A. 4. A short handoff stating what the evidence establishes and which causes remain unknown.

Do a first pass before reading the answer pages. You can calculate by hand or use a spreadsheet. Preserve your first attempt and corrections. This is self-directed practice, not an assessment of professional competence.

# Learner data and source notes

CSV column order is row_id, reference, scenario_date, account, unit, amount. Dates, account and unit are identical on every row. The compact table shows the changing fields. Headers are on row 1; A has data rows 2 to 11, B has rows 2 to 12.

A row | Ref | Amount | B row | Ref | Amount
A01 | R101 | 120.00 | B01 | R101 | 120.00
A02 | R102 | 85.50 | B02 | R102 | 85.50
A03 | R103 | 240.00 | B03 | R103 | 240.00
A04 | R104 | 60.00 | B04 | R103 | 240.00
A05 | R105 | 180.25 | B05 | R105 | 182.25
A06 | R106 | -25.00 | B06 | R106 | -25.00
A07 | R107 | 99.99 | B07 | R107 | 99.99
A08 | R108 | 75.00 | B08 | R108 | 75.00
A09 | R109 | 75.00 | B09 | R109 | 75.00
A10 | R110 | 140.00 | B10 | R110 | 140.00
 |  |  | B11 | R111 | 40.00

## Fictional source notes

S1: The exercise export log states that B04 repeats the same source item as B03, reference R103. It is an extra export row. This does not establish that any payment happened twice.

S2: The exercise receipt register lists R108 and R109 as two separate items of 75.00 FC each. Neither is a duplicate of the other.

S3: No additional source record is supplied for R104, R105 or R111. The exports reveal their differences, but do not establish which record is authoritative or why they differ. Do not invent a timing issue, fee, fraud, payment or corrective journal.

R106 is a valid signed negative practice item. Do not discard it or convert it to positive. No opening balance, bank account, tax treatment or real accounting period is supplied.

# Work through the comparison

## Step 1 Check the input

Count data rows, distinct references and numeric amounts. Check the negative value and the last row. Record the units and date scope before calculating. In your copied sheets, an amount total covers F2:F11 for A and F2:F12 for B. A spreadsheet formula is only as reliable as its range and input types; confirm every imported amount is numeric.

## Step 2 Separate repeats from matches

List references repeated within either file. For each reference, count its A rows and B rows. Only then compare one-to-one records. Keep row IDs beside each decision so a reviewer can trace the input. Equal totals do not prove equal records.

## Step 3 Explain the residue

Use the blank discrepancy-log.csv. For each exception, record A and B row IDs, raw subtotals, B minus A, observed difference, evidence and next check. Use zero only for the subtotal of an absent side; absence is not a claim that the transaction value was zero. Retain missing sides explicitly as absent.

## Step 4 Review an AI explanation if you choose

Optional prompt: Using only these fictional exports and S1 to S3, suggest a matching method and a discrepancy log. Preserve raw rows and signed amounts. Flag repeated references before one-to-one matches. Separate observed differences from hypotheses. Cite row IDs and source notes. Do not invent causes or alter records. If you calculate with code, show the code, input row counts and checks.

Check any output yourself. A prompt is not a technical permission boundary; keep practice disconnected from live records and tools. OpenAI documents that some data-analysis tasks use Python and recommends reviewing code, outputs and assumptions. Availability varies; this lab was not tested in a learner ChatGPT account.

## Step 5 Test a tempting shortcut

A proposed result says: Delete every repeated amount, ignore the negative number and add 222.00 to A to make it balance. Explain why all three instructions are unsafe. State whether a bridge that sums to the raw difference resolves every exception.

