On this page
- 1. Describe the result in one sentence
- 2. List the inputs before attaching them
- 3. Explain what a row and each important field mean
- 4. Separate known rules from decisions still needed
- 5. Start with a safe, representative sample
- A fictional brief you can adapt
- 6. Agree what a reviewable handoff means
- Use the brief, then review the method
The most useful starting point is a short brief that explains what the file represents and what you want to do with it. You do not need to solve the data problems first. You do need to identify the source, describe the intended result, and name the decisions that only your team can make.
1. Describe the result in one sentence
Start with the task the cleaned file should support. “Make this spreadsheet better” leaves the reviewer guessing. A more useful brief is: “Prepare a reviewable supplier list with our approved IDs, flag ambiguous matches, and retain the original records.”
State who will use the result, the output format they need, and whether this is a one-time handoff or a recurring process. A cleaned snapshot and a repeatable reporting workflow have different scopes.
2. List the inputs before attaching them
Record the filename, file type, export date, relevant sheet names, approximate row count, and source system. Explain whether one file is the source, an approved reference, or an earlier attempt. Identify the current version without replacing or overwriting the original.
Note hidden or filtered records, formulas, linked files, or other dependencies you know about. Say whether all rows belong in scope. If several versions disagree, name the person who can determine which source to use; do not silently pick the newest-looking copy.
3. Explain what a row and each important field mean
Specify whether a row is an order, order line, delivery, invoice line, or something else. Similar-looking rows may represent different events. Give definitions for the identifiers, dates, quantities, amounts, and status fields that matter to the result.
- Identifiers: Does 00123 differ from 123? Which punctuation or case is meaningful?
- Dates: What format is used, and what event does the date represent?
- Amounts: Which currency is used, and what does the amount include?
- Quantities: What is the unit, and are pack sizes or returns involved?
- Blanks: Does an empty cell mean unknown, not applicable, or a missing entry?
If you do not know an answer, mark it as an open question. A visible uncertainty is a better input than an invented definition.
4. Separate known rules from decisions still needed
Share approved reference lists and examples of a correct record. State which fields can be reformatted and which changes need approval. Explain any existing duplicate rule, including what makes two records the same business event.
Keep a short decision list: question, evidence needed, decision owner, and effect on the output. For example, “Can these two vendor names share one ID? Purchasing owner to confirm against the approved master.” Do not turn a name similarity into a confirmed match.
5. Start with a safe, representative sample
For an initial scope discussion, describe the file and provide a redacted or fictional example if useful. Include a normal row and a few problem rows: a suspected duplicate, an ambiguous date, a missing value, or an amount mismatch. A perfect sample can hide the work the project actually needs.
Remove unnecessary personal details and full payment information. Do not include passwords, keys, or access tokens. Check screenshots, comments, extra sheets, and filenames as well as visible cells. Agree the transfer method, permitted tools, access, and retention expectations before sending working data.
A fictional brief you can adapt
“We need a one-time review file for an order-line export. The source has 1,200 rows. Keep the original values, use our provided vendor-ID reference, and flag uncertain matches rather than merging them. Preserve SKU leading zeros. Return a cleaned view, a rule log, and an exception list. Our purchasing owner will resolve vendor questions. We will accept the handoff when every source row is accounted for, comparable totals reconcile, and unresolved items remain visible.”
This is an invented scope example. Whether a real project fits a fixed package depends on its inputs and complexity, as well as row count.
6. Agree what a reviewable handoff means
- The original file and working output are distinguishable.
- Every source row can be accounted for, including excluded or held rows.
- Comparable counts and amounts can be reconciled using the agreed basis.
- Rules, assumptions, and exceptions are documented.
- The reviewer knows which decisions are still open.
- The output format, included revision, and scope boundary are written down.
Include a deadline to consider and any dependency on a colleague’s decision. Keep new source files or new output requirements separate from corrections within the agreed scope.
Use the brief, then review the method
Download the editable project brief and cleanup preparation checklist. They use placeholder fields and can be completed locally.
The reconciliation guide covers the cleanup method itself. If you need help with the work, review the one-file cleanup pilot and its limits, then send a file description to discuss the scope.
Examples are fictional. Adapt the guidance to your records, systems, and approved policies.
Operations & Research Support