Data stories

A data story starts from a dataset rather than from a position. The analysis is shown, not summarized, and the result stands whether or not it is interesting. That last condition is what separates the form from marketing built on a chart.

What is here now

The popularity baseline was harder to beat than expected reports that tuned collaborative filtering improved on simply recommending the most-played titles by less than the evaluation’s own noise. The finding is that a simple baseline held, which is a real result and an unrewarding one.

Measuring local economic resilience is an unfinished attempt at a resilience measure computable from national statistics alone, published in that state with the limit stated: it is not fit for funding or allocation decisions.

Negative and null results are published

A story whose analysis found nothing is still published here, because the alternative is a body of work where every investigation happens to have succeeded. What collaborative filtering assumes about people and the finding above both came out of the same notebook, and only one of them is a positive result.

What every piece has to carry

The dataset and its period, the unit being counted, what the analysis cannot distinguish, and whether the numbers are described as association or cause. The standard is set out in Research and applies here without exception.