Importing a table of your own

Bring a spreadsheet of your own (an HR export, a customer list, a supplier ledger, a payments file) into a case beside what the lookups found, and beside phone records and device extractions (Phone records and device forensics). The rows become the same kinds of things and links the OSINT sources write, so a company in your file and the same company from Companies House are one node.

The console steps are in Import records; this page is the detail.

The steps

  1. File. CSV (comma, semicolon or pipe, read off the header line), TSV, or JSON lines (one object per line; nested keys read as address.postcode). The first 50 rows are shown; under each heading is what most of its values look like (emails, phone numbers, company numbers, dates, IBANs…). Nothing is recorded yet.
  2. Columns. Say what each row describes and which column fills each field. The console guesses from the headers and the shape of the values; correct it. Beside every column is how many values parse ("41 of 45 parse, 4 do not (e.g. "tel. in reception")"). Properties ride along as identifiers: email, phone, IBAN, bank account, date of birth, website, job title, free text.
  1. Links. Optionally a second thing per row and the link between the two (“employee → employed by → company”, “customer → uses → handset”; the console lists the link types), a date column (from, to) giving the link its dates, or “each row is an event” (a payment, a meeting): one event per row, with both things taking part.
  1. Source. The chain of custody: exhibit reference and where it came from (both required), who handed it over, and your grades. Source reliability A to F and information credibility 1 to 6 (with the reason in your words, required when you grade).
  2. Check and import. The dry run in plain words ("42 people, 4 companies, 42 links; 6 rows skipped"), why each skipped row was skipped with the first row numbers and example values, how many things already match the case, and "save this mapping for the next file". Then import.

Afterwards: what you can do next (run lookups on the imported companies, check sanctions for the people and companies imported; the first 10 are run, progress is in the activity bar).

What is a thing

Kind Key fields
Company or organisation company number + country (default GB), or LEI, or just the name
Person full name (two words at least), date of birth
Domain domain, website or URL
Email address email
Phone number phone, to E.164 with the default country
Postal address up to three lines, town, postcode, country
Vehicle registration + country, or VIN
Handset IMEI (check digit verified)

Keys are built the same way for every source, so 1234567 and 01234567 are one GB company. A bank account is not a thing of its own: map an IBAN (checked) or an account number as a property of the person or company it belongs to.

A value that does not key is refused, never repaired by guessing. If a key field of a row does not parse (a one-word person name, a company number that is not valid for its registry, a date of birth that is not a real day), the row is skipped, counted, and the reason is shown. A property that does not parse is left out and counted in its column; the row is still imported. A blank second thing imports the row without the link. Repeated rows fold into one thing per key within the import.

Evidence and grades

Every import keeps two things in the evidence store, both identified by SHA-256: the file exactly as supplied, and a custody record (exhibit, file, who imported it and when, the mapping, your grades, counts). Every step cites both. The stored file can be re-hashed to check it.

  • The source grade is recorded as your decision on the import's source, with your reason, like any source grade; the case shows it where it shows grades.
  • The information grade sets the confidence every link of the import is asserted with: 1 confirmed, 2 probable, 3 possible, 4 to 6 unverified. The grade the case shows is computed from it and from corroboration by other sources.

Purpose gate

An import is refused until the case has a recorded purpose and lawful basis (declare it in the “Before you look up people” panel; see People investigations) when it maps a person, phone number, email address or postal address as a thing, or a phone, email, IBAN, bank account, date of birth or address as a property. A file of companies, domains, vehicles and handsets alone is not gated. Hold the authority the data was obtained under and import only what it covers; Viokit records what you state, you are the controller.

Matching what the case already holds

  • Identifier matches are automatic: the same key is the same node, so a company number in your file joins the Companies House company already in the case. The dry run says how many.
  • Name matches are proposals only. A person known by name, or a company known only by name, that the person, organisation or address matching would match is recorded as a pending merge proposal (see the Matches panel), never merged. Namesakes with different birth months are never proposed. Names alone never make one person.

Templates

“Save this mapping” stores the mapping as a template of the open case. Pick it on the next file of the same shape (“Start from a saved mapping”); a file that lacks a column the template reads is told so.

Limits

No Excel files (save as CSV). Row numbers are the Nth data row, not the line in the file. Dates without a zone are read as calendar days (UTC). Only the first ten companies or names are looked up from the "what next" buttons; run lines of investigation from the case for the rest.

Docs

↑↓ to moveEnter to openEsc to close