Jeff Su: Clean a Sheet Without Losing Distinct Records

The Lesson

A sheet can look cleaner while becoming less accurate. Two rows with the same name may represent two real registrations. An empty cell may mean someone still needs to answer a question. Before changing either, decide what one row is supposed to represent.

In his official Google Sheets tips article, Jeff Su describes data-cleaning tools alongside checks such as counting occurrences and viewing cell history. His official resources page identifies his focus on helping professionals with productivity and organization. That combination gives us a useful starting point: make a change, then inspect what it did.

A cleanup tool cannot supply the meaning of your records. Google’s current documentation explains that Remove Duplicates uses the columns you select. If you select only a name column, you have chosen a different comparison from one that includes a registration identifier and session. Decide the comparison before using the tool.

For this practice, one row represents one registration. The fictional entry identifier distinguishes registrations; a name does not. Preserve the untouched source, work on a separate copy and keep unresolved questions visible. The goal is a result another person can inspect, including the choices you made.

Reflection

• What does one row represent: a person, an order, a registration or something else?
• Which field distinguishes records, and who established that rule?
• Am I removing an accidental repeat or erasing a separate event?
• What would I need to ask before resolving a blank or conflicting value?

Practice

Original SelfGrowthVideos exercise: This fictional registration rehearsal is our application of the checking idea. It is not Su’s dataset, worksheet or an endorsed cleanup method. Use a spreadsheet you already know or work through the rows on paper.

Create these five rows and save an untouched version named Raw. The brackets describe a blank; they are not text to enter.

1. Write the rule. In this sample, each entry identifier belongs to one registration. Matching names alone do not establish a duplicate. If the same identifier had conflicting details, you would need to resolve the conflict before deciding what to keep.
2. Make a working copy. Name it Clean. Remove the ordinary leading and trailing spaces around the first name. Check that its identifier and session remain unchanged.
3. Compare candidate repeats. Keep both R01 and R02: their identifiers and sessions differ. Inspect both R03 rows across all three columns. They are an exact repeated registration under the sample’s rule; retain one and record why you removed the other.
4. Preserve the unknown. Keep R04 and its blank session. Add a review note asking which session belongs there. Do not invent an answer or remove the registration to make the table look complete.
5. Reconcile the result. You should have four rows, with identifiers R01, R02, R03 and R04. Compare every retained row with Raw. The expected count is a check, but a count alone cannot show that you kept the correct records.
6. Write a handoff note. State the row rule, the space adjustment, the exact repeat removed and the unanswered session question. Keep Raw, Clean and this note together so the changes can be reviewed.

Review

Ask another reader to explain why the two Ari registrations remain. If they cannot tell from the files and note, improve the explanation before making more changes.

Now imagine a new R03 row with Afternoon instead of Morning. Would your procedure flag conflicting details, or silently remove one version? Revise the review step so that a disagreement stays visible. This sample tests your process; it does not establish that an unfamiliar real dataset follows the same identifier rule.

Go Deeper

Visit Jeff Su’s profile and the Learning & Career AI track. Compare this record-level decision with giving different inputs a common shape and leaving room for uncertainty in triage rules. Return to Virtual Assistance & Administration for related practice topics.