Clean the keyword column. Keep the data you still need.
A good import cleanup removes formatting noise and duplicate keyword rows without flattening a useful spreadsheet into a text list.
Map first. Clean second.
Start by choosing which column contains the keyword and which supporting columns must survive. Then normalize only the keyword values. Search volume, list names, language, intent, or custom tags may be needed by the receiving product even when the keyword text itself needs cleanup.
- 01 / Source
Find the real table shape.
Check whether the source uses commas, tabs, semicolons, pipes, or one item per line. Quoted CSV fields can contain commas, so splitting every comma blindly can corrupt a keyword or supporting field.
Before cleanup: identify the separator, header row, keyword column, and columns required after import.
- 02 / Mapping
Name the keyword column explicitly.
Do not assume the first column is the keyword. Crawler and SEO exports often start with an ID, URL, date, or project name. Preview several rows before choosing the field and keep only the additional columns with a clear downstream job.
- 03 / Cleanup
Remove presentation noise, not meaning.
Whitespace, list numbering, blank separator rows, and accidental duplicate capitalization are usually safe cleanup targets. Punctuation inside a phrase, accents, hyphens, numbers, and slashes can carry meaning. Do not remove them with a generic “special character” rule.
Safe default: trim edges, collapse repeated spaces, remove obvious bullet prefixes, and stop for review when duplicate keywords carry different supporting values.
- 04 / Review
Make every loss countable.
Compare input rows, valid keyword rows, duplicate rows, metadata conflicts, noise removals, and final rows. Identical duplicates can collapse safely. If the same keyword has different volume, intent, language, list, or custom values, choose which row wins or merge the distinct values deliberately.
The workbench core contract exercises multi-column mapping, conflict handling, noise removal, and the Contextter-compatible keyword header against the same production module used in the browser.
- 05 / Export
Choose the next system, then the format.
Use one keyword per line for a plain queue, CSV when columns must survive, and JSON when another script needs explicit field names. The Contextter CSV option names the selected field
keywordand retains selected supporting columns for manual mapping; it does not upload or import the file.
Two duplicate keywords; one real metadata conflict.
The synthetic CSV contains two identical Keyword Clustering rows and two Technical SEO Audit rows with different volume, intent, and list values. The first pair collapses. The second pair blocks export until the recipe selects a merge.
- Decision
- Merge distinct retained values with a visible separator.
- Expected failure
- Removing the saved resolution makes the regression report one unresolved conflict.
- Boundary
- No volume or intent is treated as more accurate.
Format and product contracts.
- RFC 4180: Common Format and MIME Type for CSV Files
The baseline reference for comma-separated records, headers, and quoted fields.
- DevAwesome workbench core contract
The executable local assertions behind mapping, cleanup, deduplication, and export behavior.
Cleanup cannot choose your keyword strategy.
The workflow does not calculate volume, intent, difficulty, clustering, or ranking value. It prepares a reviewable input for a separate analysis.
One result should be easy to audit.
Keep the raw source until the destination has accepted the import. A clean export is useful; it is not a replacement for the original evidence.