Services / Data cleaning & preparation
Data cleaning that leaves an audit trail, not just a tidier file
Most analysis problems are data problems. We clean the dataset, document every decision, and hand back both the cleaned file and a record of what changed — so the cleaning itself can be defended, not just the result.
- Cleaned, analysis-ready dataset
- A written log of every change
- Original file left untouched
- Syntax so cleaning can be rerun
Data cleaning is the process of making a dataset fit for analysis: resolving structural problems, diagnosing and treating missing values, investigating outliers, recoding variables and scoring scales. Done properly it produces a change log and a reusable script, so every decision can be audited and repeated.
What is included
What you get with data cleaning & preparation
Structure and merging
Wide to long, joining files, resolving mismatched IDs and inconsistent keys across sources.
Missing data handled properly
Diagnosed before it is treated. Listwise deletion, imputation or a caveat — chosen deliberately and explained.
Outliers investigated
Flagged, checked and only removed where there is a reason. Deleting inconvenient points quietly is how analyses fall apart under review.
Recoding and derived variables
Scales scored, reverse items handled, categories collapsed and derived measures built to your definitions.
A rerunnable script
Every step in syntax, so the cleaning can be repeated exactly when more data arrives.
A change log you can show
A plain-language record of what was changed and why, suitable for a supervisor, reviewer or auditor.
Who it is for
Two kinds of client, one standard of work
- Survey exports that are unusable as downloaded
- Missing values scattered through the dataset
- Scales that need scoring and reverse-coding
- Merging data collected in several waves
- Records from several systems that do not match
- Years of inconsistent manual data entry
- Preparing data before analysis or a dashboard
- An audit trail for anything reported externally
The method
Why cleaning is the analysis
Most problems presented to us as analysis problems are data problems. A model cannot repair a dataset in which missing values were deleted without diagnosis, scales were never reverse-scored, or three systems disagree about who a record refers to.
Cleaning is a sequence of decisions, and each one shapes the result. What makes cleaning defensible is not that it was thorough but that every decision was recorded, justified and reproducible.
The hardest part
Missing data, handled deliberately
Deleting incomplete cases is a decision, not a default — and it is the wrong one whenever data is not missing completely at random, which is most of the time. People who drop out of a study differ from people who stay.
| Pattern | What it means | Usual treatment |
|---|---|---|
| MCAR — missing completely at random | Missingness unrelated to anything measured or unmeasured | Listwise deletion is defensible, though wasteful |
| MAR — missing at random | Missingness explained by other observed variables | Multiple imputation, or a model that uses all available data |
| MNAR — missing not at random | Missingness depends on the unobserved value itself | No clean fix. Sensitivity analysis and an explicit limitation |
The pattern is diagnosed before anything is treated, because the diagnosis determines the treatment. Whichever route is taken, it is stated in the write-up rather than left implicit — reviewers ask, and “cases with missing data were excluded” invites the follow-up question.
What cleaning covers
The rest of the work
Structure and merging
Wide to long, joining files, resolving mismatched IDs and inconsistent keys across sources and collection waves.
Outliers investigated
Flagged and checked, removed only with a reason. Quietly deleting inconvenient points is how analyses fail under review.
Recoding and scoring
Scales scored, reverse items handled, categories collapsed, derived measures built to your definitions rather than guessed.
Distribution checks
Identifying where transformation or a different model class will be needed, before the analysis rather than after.
Consistency and validation
Impossible values, out-of-range responses, duplicate records, contradictory answers within a case.
A rerunnable script
Every step in syntax, so the cleaning can be repeated exactly when more data arrives or an error is found.
The deliverable
What you get back
Anyone can produce a tidier spreadsheet. What a supervisor, reviewer or auditor asks for is the record of what changed between the raw data and the numbers in the report — and that is the part most cleaning work never produces.
Pricing, and why we look first
The cost depends entirely on the state of the file, so we assess it before quoting and there is no charge for that. A well-structured export with a few missing values is a different job from twelve years of manual entry across three systems, and quoting blind would mean either overcharging you or discovering the truth halfway through.
Practicalities
Formats, timescales and confidentiality
If the file turns out to be in better shape than expected, the quote goes down rather than staying where it was. The assessment exists to price the job accurately, not to justify a number decided in advance.
How it works
Three steps, entirely in writing
Send what you have
Attach the data, or just describe the project. No account, no mandatory call, and a five-field form rather than a fifteen-field one.
Approve a fixed quote
Usually within one working day, naming the statistician assigned, the deliverables and the deadline.
Receive the work
Output, syntax and a plain-English explanation — with follow-up questions answered at no extra cost.
Client feedback
What clients say about this work
“They cleaned a large dataset quickly and identified several issues that would have affected my analysis.”
“Excellent data preparation service. Missing values, duplicates, and inconsistencies were all handled professionally.”
Proof
This work, on real projects
Clients are not named and no identifying detail is published, so what is described is the statistical problem — which is the part that shows whether a consultancy knows what it is doing.
Principal Component Analysis and K-Means Clustering for Customer Segmentation
Finding stable segments in 90,000 customer records with no predefined categories and heavily correlated behavioural variables.
Manufacturing · Generalised linear mixed modelsGeneralised Linear Mixed Models to Identify the Drivers of Production Defects
A quarter of a million production records across six lines and two years, where univariate analysis could not separate interacting process variables.
Public services · Time series forecastingSeasonal ARIMA Forecasting with Intervention Analysis for Resource Planning
Twelve years of monthly demand with seasonality, trend and structural breaks — where simple smoothing methods failed consistently.
Your statistician
Who does this work
Member of the Royal Statistical Society · Member of the British Computer Society
Machine learning and predictive modelling for high-dimensional data across healthcare, finance and education.
This page was reviewed for statistical accuracy by Tara Sutaria on 14 August 2026.
Common questions
Answers before you ask
Will you change my original data?
Never. The original file is left untouched. You get a cleaned copy plus the script and log describing exactly what differs.
How do you decide what to do about missing data?
By diagnosing why it is missing first. The pattern determines the treatment, and whichever approach is used is stated in the write-up rather than buried.
What formats can you work with?
SPSS, Excel, CSV, Stata, R, Qualtrics and SurveyMonkey exports, and most database dumps. Ask if yours is not listed.
Can you clean data you did not collect?
Yes — most of what we clean was collected by someone else, sometimes years earlier.
How much does it cost?
It depends entirely on the state of the file, so we look first and quote after. There is no charge for the assessment.
Related
Often needed alongside this
SPSS data analysis
The right tests, annotated output, reusable syntax and APA-ready tables.
From £95Statistical consultancy
Design, analysis and reporting for charities, universities, the NHS and business.
From £950Dissertation statistics help
Analysis of the data you collected, explained test by test so you can defend it.
From £95Send the data. Get a fixed quote.
Attach your dataset or just describe the project. A named statistician replies with a price and a deadline, usually within one working day.