About Datanemics

Datanemics is a crash course in data work, a library of datasets to practise on, and a set of tools that run in your browser. It is written by practitioners across nine disciplines, and attributed by practice area rather than by person — every lesson and post carries the discipline it came from.

What is here#

The crash course is 36 lessons across ten modules, ordered the way the work actually happens: get the data, discover it is wrong, model it, measure something, tell someone, keep it running. Every lesson ends with an exercise you can run against a dataset on this site.

The pattern library is 24 named problem shapes — the things you look up rather than read through. Each names the version that looks right and is not, then the version that holds, and points at a dataset here to try it on.

The dataset library is 14 synthetic datasets, roughly 73,000 rows. Each one is built around a specific analytical trap — informative missingness, right-censoring, a relationship that reverses when you disaggregate, five date formats in a single column. The dataset page names what is buried in the data and suggests what to try.

The tools are seven browser-based instruments: a SQL engine, a chart builder, a data explorer, a regression lab, a distribution sandbox, a correlation explorer and an experiment calculator. They run against the datasets here, or against a CSV you drop in — which stays on your machine, because this site has no server to send it to.

The articles are longer pieces on how the tools work, where standard advice breaks down, and what to do instead.

The field notes run one a day from May 2023. Most of them profile something concrete on this site — a column, a key, a correlation, a segment — and every number they quote is computed from the dataset it names, so you can check any of it against the CSV on the dataset page. Some are notes on the course, the patterns, or how this site is put together. The daily ones are composed from measured statistics rather than written one at a time; the longer pieces in articles are not.

Why synthetic data#

Because the lesson can be put in the data. A dataset built to demonstrate informative missingness can guarantee the missingness is informative; a scraped one cannot. It also means no licence trap, no personal data in a tutorial, and no numbers that shift under you — every file is generated from a fixed seed, so the row a lesson references stays that row.

What you can do with it#

• Course, articles and field notes — CC BY 4.0. Teach from it, translate it, adapt it, put it in your company's onboarding, including commercially. Keep the attribution.
• Datasets — CC0 1.0. Public domain, no attribution required.
• Code — MIT. The generator and every tool.

If you translate a module, we would like to link to it. That is a request, not a condition.

What this is not#

It is not a bootcamp: there is no cohort, no mentor, no certificate. It is not comprehensive — ten modules will not make anyone a senior engineer, but they will give you the shape of the whole field so that when you go deep on one part, you know where it sits. And it is not neutral. We have opinions, we state them, and where one is contested we say so.

How it is built#

A static site, generated by about 900 lines of Node with one dependency. Every page renders without JavaScript — the whole course, every article, every dataset page. JavaScript adds search, the theme toggle, progress tracking and the tools, and nothing breaks without it.

Charts follow a fixed set of rules enforced in code rather than by convention: one y-axis and never two, colour tied to the entity rather than its rank, a legend whenever there is more than one series, and a data table under every chart. The categorical palette was validated for colour-vision deficiency with a script, against both the light and dark chart surfaces.

There are no trackers, no cookies and no analytics. Course progress is stored in your browser and never leaves it.

Scheduled posts#

Posts carry a date. Anything dated in the future is written but not published — it gets no page, and appears in neither the sitemap, the feed, nor the search index. A scheduled build runs every morning, so a post dated three months out appears on its own three months from now.

Corrections#

The course contains errors and always will. When one is found, the fix goes in the page with a note about what changed rather than being edited silently; the corrections post collects them. The repository is on GitHub and the issue tracker is open.