List datasetsFind the Data USA dataset that holds the numbers a question needs. Search by subject and every dataset whose name, title, topic, survey or agency mentions it comes back, each with the measures it reports and the levels it can break them down by. This is the step that turns 'how many people live in Texas' into a dataset name the other actions can read. Every action here reads: nothing changes anything. No API key required.
List topicsSummarize everything Data USA publishes by the subjects it covers: how many datasets sit under each topic, the subtopics inside it and the agencies the numbers come from. The step to run first when the question is whether the site has anything on a subject at all, before searching for a dataset by name. No API key required.
Get a dataset's schemaGet one dataset's measures, dimensions and levels: the numbers it reports, what each one is aggregated by, and every column those numbers can be broken down across, with how many distinct values each column holds. The levels also come back as one flat list, because that is what a query's drilldowns are written in. Read this before writing a query rather than guessing at column names. No API key required.
List a level's membersList the distinct values of one column of a dataset, each with the key a query has to be written in. Filters are written in keys rather than names, so this is the step that turns 'Alabama' into '04000US01'. Search to find one member in a level that holds thousands, such as every county in the country. No API key required.
Get dataRun a query against a Data USA dataset and get the rows back. Name the dataset, the columns to break the numbers down by and the numbers themselves, then narrow the answer: to particular members, to everything but particular members, to rows above or below a threshold, to the most recent periods the dataset holds. Sort it, page through it, and pull in each row's parent so a county row also says which state it is in. A measure is aggregated over everything it is not broken down by, so include the time level unless you mean the total across every year. No API key required.
Get the top rows per groupGet the highest or lowest few rows for each group, in one request rather than one per group: the three best-paid occupations in every state, or the single year each state's population peaked. Every column the ranking names has to be one the query already asks for, which is checked before the request goes out so a mistake comes back as a sentence rather than a rejected query. No API key required.
Get a measure over timeGet how one number has moved: the series period by period, oldest first, each with the change from the period before, plus the change across the whole window, the average, and the highest and lowest periods. Name a member to narrow it to one place or one category, by its name or by its key. The most recent periods are the ones returned, so a step can say 'up 2.1% on last year' rather than quoting a number with nothing to compare it against. No API key required.
Compare membersCompare one number across several places or categories, ranked biggest first with each one's share of the total and how far it is behind the leader. Members can be given by name rather than by key, because the level's members are read and the names translated, and anything that matched nothing is reported as missing rather than quietly dropped. The most recent period is compared by default, because a measure added up across every year at once is a number that looks right and means nothing. No API key required.
