Keep a baseline
Retain the original data separately. Make transformations in a working copy so decisions can be checked and, where possible, reversed.
RESEARCH & DATA
Data cleaning & preparation
Prepare research data with decisions you can trace. We preserve the original file, agree transformation rules and record what changes before analysis begins.
Request a research quoteWhen this helps
Your survey export, spreadsheet or combined dataset contains inconsistent codes, missing values, repeated records or variables that need to be prepared for analysis.
THE DETAIL THAT MATTERS
Not every outlier is an error.
An unusual value may be a recording error, an expected extreme or an observation that changes the interpretation. We check its source and context before deciding what to do. Exclusions need a reason, not just a threshold.
- 01Original data
- 02Checks
- 03Decisions
- 04Prepared data
- 05Change log
What the work involves
Review the structure
Check identifiers, record counts, duplicates, variable types, labels and allowed values. Repeated observations are not automatically duplicate records.
Investigate missingness
Distinguish unanswered items, skipped questions and special codes. Review patterns before agreeing exclusions, recoding or other treatment.
Document transformations
Record recodes, derived variables, merging rules and exclusions. Check the prepared export against those decisions.
What you receive
Deliverables are agreed in your quote.
- A prepared dataset in the agreed format.
- A variable list or updated codebook.
- A change log with rules, reasons and affected records.
- Data-quality notes, unresolved questions and analysis implications.
WHAT THE WORK LOOKS LIKE
Decisions you can follow.
A preparation decision should leave a record. This extract shows the kind of reasoning a change log can capture.
ILLUSTRATIVE EXAMPLE
This is an example of the format, not a client project. Your quote specifies the documents and outputs included in your handover.
DATA PREPARATION / DECISION LOGExample 01
- Observation
- A response value falls outside the questionnaire’s allowed range.
- Check
- Compare the value with the source export and coding instructions.
- Decision
- Correct only if the source supports a correction. Otherwise record the unresolved issue.
- Record
- Keep the original value, the reason for the decision and the working-file version.
What to prepare for an enquiry
Share your research question, stage, data type and deadline. No test selection is needed.
Please leave out sensitive records. File sharing is agreed after the initial review.
QUESTIONS
Questions about this service
Will you delete outliers?
Not automatically. We investigate unusual values and agree defensible rules. A valid extreme observation may need to remain in the analysis.
How do you handle missing values?
We review what the codes mean and how missingness is distributed. The approach depends on the study and proposed analysis; there is no single rule suitable for every dataset.
Do you alter my original dataset?
We preserve it separately and work on a copy. Your handover identifies the version used for analysis and the changes applied.
Will I receive a record of changes?
Yes. The agreed preparation work includes a record of important transformations, exclusions and issues that still need attention.
NEXT STEP
Discuss this research need.
We’ll review your brief and confirm the scope and fee.