Skip to main content
A target’s folder is everything PodData holds tied to one of your team’s targets, as files: a part for each company, the target first, then each confirmed rival, its listed parent and its comparables.

1. Read the index

Each part has its role (target, rival, parent or comparable) and its company, and lists its sources. For each source:
  • records and signals: one entry a kind, each with its count, its first and last day (from, to), file, the CSV’s address, and rows, the same rows as JSON. Where file is null, held says why.
  • documents: how many, and the first and last day filed.
  • pages, for the company’s own site: the pages read, with their file.
For the target, the index adds:
  • quotes: two files, what its people said (said) and what others said of it (named).
  • licensed: the figures your team’s own data licences read, by vendor, as JSON.
  • reports: your team’s reports on it, each with its JSON address (rows) and its zip (file).
  • not_held: every source that holds nothing for it, with where it stands and why.
  • download: the address of the zip.
The MCP tool target_folder gives the same index.

2. Fetch each file

Fetch each file by its own address, with the key. A file holds one source and one kind. A file longer than one answer holds stops at its row limit, says so in its last row, and gives the cursor for the rest in the X-PodData-Next header: ask again with cursor set to it.
Most review and retail platforms’ reviews come as a file. Where one is held back until its terms allow handing it on, file is null and held says why: read its rows as JSON, a page at a time, and save them as JSON.

3. Or take the zip

It holds the folder’s files, a part a folder, with index.json, coverage.csv and a README.txt that lists each file with its address, and what it leaves out and why. A file that would take it past its row limit is left out, and the README gives its address. PodData’s own copies of originals aren’t in it: each is read on its own with GET /documents/{id}, under the key’s daily limit. The diligence skill’s puller fetches the index and every file in one command: see A diligence, start to end.

Each file’s columns

A file is at most 10,000 rows; the zip holds at most 100,000 in all.

A source’s records

Then a column for each of the records’ own fields, by its name: a registration’s npi, a licence’s expires.

A source’s signals

Then a column for each of the signals’ own counts.

A company’s documents

Every cell keeps to the rule of the rest of the API: at most 80 words in a row from a source, and no reviewer’s or poster’s name, since none is held.