Keep leading zeros and long IDs.

Postal codes, product codes, tracking references, account numbers, and long transaction IDs usually need exact string preservation. Quantities, prices, and measured durations usually need numeric types. Make the choice from the column's meaning rather than the appearance of its first few rows.

Create a simple schema note: customer_id is text, order_date is a date under a named convention, and net_amount is a decimal in a named currency. A column containing 00127 and 127 may represent two different accounts. Do not collapse them without evidence from the source system.

In Excel, import through Data and Text/CSV, enter the transformation step if necessary, and explicitly set identifier columns to Text before loading. Inspect automatically generated type-conversion steps: a later Text step cannot recover digits a previous Number step already discarded.

Recent Excel versions expose controls for some automatic data conversions. Availability depends on the version and build, so a documented Text import remains a useful repeatable method. Formatting an empty destination column as text may help manual entry, but double-clicking a CSV can bypass the import contract you intended.

Suppose a known five-character postal code 00127 becomes 127. Padding to five characters can restore the displayed code only if the authoritative schema guarantees that length. The same repair would be wrong for a variable-length customer ID where 127 and 00127 are distinct.

Excel numeric precision is limited to 15 significant digits. Once a longer reference has been converted and rounded, adding zeros or switching to Text does not recover the original suffix. Scientific notation may be only a display choice, but the full stored value must be checked. Reimport an untouched source to distinguish display from irreversible conversion.

Use a synthetic sample with 00127, 127, a reference longer than 15 digits, and a code containing an E. Compare the imported values character for character against the source. Include blanks and intentionally repeated keys so validation does not assume every row is unique.

In Excel, EXACT can compare two text strings, but inspect the values before treating any mismatch as corruption. Leading or trailing spaces may be source errors, meaningful padding, or import artifacts. Keep the original column and place any normalized comparison key in a separate field.

Imagine 100 orders totaling $8,000. Ten orders have damaged customer references and fail to match the customer table. A left join still returns $8,000, so a revenue-total check passes. Regional revenue is nevertheless incomplete because those ten orders have no reliable region.

• Reconcile row counts and amounts before and after the match.
• Count unmatched orders and the amount attached to them.
• Confirm the lookup table has one row per intended customer key.
• Keep unmatched records visible instead of assigning them to a guessed customer.

The spreadsheet merging guide explains how duplicate lookup keys can multiply rows. Preserving identifiers prevents one class of failure; it does not prove the relationship between two tables is correct.

CSV carries characters but no durable column-type declarations. A recipient can reopen your carefully prepared file and repeat the same conversion mistake. Send the import instructions with the CSV, or share a workbook that preserves text-typed cells when the recipient's workflow allows it.

For a Data Analysis App project, review identifier columns and matching limitations before interpreting customer-level findings. Start with a copy of the export, retain the untouched original, and treat unexplained missing keys as a blocking data issue rather than a small formatting inconvenience.