Editorial standards

Figures attributed to the Nova first-party panel are computed from a fixed panel of several hundred Amazon sellers who have connected their Amazon accounts to Nova. We publish ratios, shares, rates and index values. We do not publish absolute revenue or spend, and nothing we publish identifies a seller, a brand or an ASIN.

Every Nova figure carries the period it covers and the basis on which it was measured, and every one of them links to the panel methodology, which sets out what is measured, what is excluded and what is never published.

Nova figures live in versioned data files and are rendered from those files. A number shown on a page is the number in the dataset, not a number typed into the copy.

A statistic we did not compute carries a named source and a working link to the specific page that backs it, not to a homepage or a press index. Where a source states a period, we state the same period.

If a claim cannot be sourced, it is cut. We do not soften an unsourced claim with words like "many" or "most", and we do not cite an aggregator that is itself repeating an unsourced figure.

A Nova figure is never presented as third-party research, and a third-party figure is never presented as ours.

Where an article walks through numbers to show how a calculation behaves, that is a worked example. It is labelled as one, and the arithmetic is shown so you can substitute your own inputs.

A worked example is never written as a customer story. We do not publish invented case studies, invented customer names, or results attributed to a customer who did not report them.

Timebound articles and news are reviewed every 90 days. Evergreen articles are reviewed every 180 days. The rule lives in code, and the review queue is generated from it rather than tracked by hand.

A review means the page was read, its claims checked against current sources, and its links tested. The date on a page records a review that actually happened. We do not bump a date to make a page look fresh, so a page that needed no change keeps its older date on purpose.

Where a page shows a next review date, that date is derived from the cadence rule and the last real review, not from the day you load the page.

If a figure, a source or a calculation on this site looks wrong, tell us and we will check it. Email [email protected] with the page URL and what you think is wrong, or use the contact form.

When a published figure turns out to be wrong, we correct the page and move its review date, because a correction is a real review. Where the error changed the meaning of the page, we say what changed.