Standardisation
Agree formats, vocabularies, dates, units and allowed labels, without erasing meaningful differences.
03 / Data cleaning & preparation
Cleaner data.
Clearer decisions.
Prepare datasets for analysis, migration or everyday operations. We assess inconsistent formats, possible duplicates, missing values and invalid records, then work to agreed transformation and quality-control rules.
What a data cleaning project addresses
Data can be technically readable while still producing unreliable results. Different spellings split categories, leading zeroes disappear from identifiers, quantities use incompatible units, and repeated names create misleading duplicate counts. Missing values may mean “unknown”, “not applicable”, or simply “not recorded”. These are not always interchangeable.
We profile the information and agree how it should behave for its intended use—analytics, migration, cataloguing, reporting or integration with another dataset.
Duplicates and matching
Review repeated records and identify likely duplicates, separating legitimate repeats from true duplicate entities.
Validation and quality
Check required fields, identifiers, relationships, ranges and data types; produce a documented exception list.
Agree cleaning rules before changing data
Good data cleaning is reproducible. A documented rule might convert an explicitly labelled weight of 0.8 kg to 800 g, preserve 000123 as a textual identifier, or map two spelling variants to an agreed controlled vocabulary. A date such as 03/04/2025 cannot safely be interpreted without a known source convention.
Where a value cannot be verified, leave it unresolved or record the decision process. Do not silently manufacture dates, infer legal rights or overwrite conflicting source values. The goal is a dataset whose quality and limitations can be explained.
An accountable preparation workflow
- 01
Profile the data
Review structure, field completeness, outliers, identifiers, vocabulary and likely duplicates.
- 02
Agree acceptance rules
Define what is valid, which values can be mapped, which fields must be preserved and when human review is required.
- 03
Apply and reconcile
Perform the agreed transformations, retain source references and compare before/after quality checks.
- 04
Deliver with evidence
Provide the prepared export and relevant mappings, decisions, exception records and quality summary.
The result should be usable—and understandable
Typical outputs may include a standardised CSV or database export, a field dictionary, a list of candidate duplicate matches, mapping tables, an unresolved-issue register and the tests performed. The exact deliverables depend on the use case and input quality.
Cleaning does not recover information that never existed in the source. It cannot guarantee zero errors without a defined and verifiable acceptance criterion. If multiple sources also need a common schema, our data integration service can address that connected requirement.
Common questions
Do data cleaning and data cleansing mean the same thing?
They are commonly used for similar quality-improvement work. The important part is agreeing what errors or inconsistencies need attention and how results will be checked.
Can you remove duplicate records?
We can review candidate duplicates and apply agreed resolution rules. Similar-looking records are not always the same entity.
Can missing information be filled in?
Only where a verified source or agreed rule supports the value. Otherwise unknown or missing data should remain explicit.
Can you prepare data for a system migration?
Yes, preparation and validation can be scoped for a destination system. Operation of that system is a separate engagement unless agreed.
Do we need perfect source files?
No. Initial profiling identifies the issues and helps determine a realistic process. Begin with a description rather than confidential files.
Start with a description
Make your data more usable.
Describe the data's format, approximate volume, quality problems and intended outcome. We can define a practical next step from there.
No confidential dataset upload needed.
Scope and next steps come first.
Enquiry submitted
Thank you for getting in touch.
Your enquiry has been accepted for sending. We will review your message and get back to you.