# data · data.sql
Data Cleaning
Messy datasets turned into an analysis-ready file, with every step logged.
A cleaned dataset with no record of what was done to it is hard to defend when a marker asks why a column changed shape. We log every decision, flag outliers instead of quietly dropping them, and hand back a script that reproduces the clean file from the raw one — so the process is as gradable as the result.
What you get
- Every cleaning decision logged — what was dropped, and why
- Missing values handled with a method your brief can justify
- Outliers flagged, not silently removed
- Before/after summary statistics included
- Delivered in the format your analysis tool expects
- Reproducible script provided, not a one-off manual clean
What to send first
- The dataset (or a sample of it) and where it came from
- The questions the analysis has to answer, or the model's target
- Required tools — SQL dialect, Python libraries, R, Excel version
- The report format: notebook, PDF write-up, dashboard or slides
How it runs
- 01
Send the spec
The brief, the rubric, any starter code and the versions your marker uses. A photo of the PDF is fine to begin with.
- 02
Scope and quote in writing
We read it properly, ask what is unclear, and confirm a fixed price and delivery date. A small advance confirms the order.
- 03
Build against the brief
Work is tracked task by task against the spec. Longer projects are split into milestones you can review as they land.
- 04
Tested handover
A repo that installs from a README, run against your test cases, with the write-up if your rubric asks for one.
- 05
Fixes and walkthrough
One round of fixes if a marker flags something, and a walkthrough of how it works so you can explain it.
Data Cleaning: common questions
The notebook or scripts that produced every number, so the analysis can be re-run end to end. Results pasted into a report without the code behind them are hard to defend.
No. Data is used only for your task, never shared or reused, and deleted on request once the work is signed off.
It depends on the scope of the brief, the stack and the deadline. Send the spec on WhatsApp and you get a fixed price and delivery date in writing before any work starts — no hourly meter.
Turnaround depends on scope and is agreed with the quote. Tell us the real deadline, including the timezone, and we will say honestly whether it is workable before you commit.
// related
Often paired with data cleaning
Database Design
Normalised schema design from an ER diagram through to a working database.
data.sqlSQL Query Work
Query sets solved and explained line by line, from joins to window functions.
data.sqlData Pipelines
ETL pipelines built to ingest, clean and load a dataset the way your brief specifies.
./send-brief
Send the raw file.
Paste the data cleaning brief into WhatsApp — the message already names the service. Scope and price come back in writing.