Merge CRM Contact Exports Into One Unified List

Drop all three exports onto Step 1 at once. The single rename map normalises the merged file; one pass produces the unified list.

Step 1 (csv-merger, union) — drop together, in preference order:
  1. salesforce.csv  (Id, FirstName, LastName, Email, Phone, AccountId, Title)
  2. hubspot.csv     (Record ID, First Name, Last Name, Email, Phone Number, Company)
  3. pipedrive.csv   (Person ID, Name, Email, Phone, Organization)
  -> merged.csv = union of all columns

Step 2 (csv-header-rename) on the merged file:
  FirstName/first_name -> First Name
  LastName/last_name   -> Last Name
  EmailAddress/email   -> Email
  Account Name/Organization/CompanyName -> Company
  AccountId -> Account ID

Step 3 (csv-case-converter, lower): emails (and all cells) lower-cased
Step 4 (csv-deduplicator, Email): one row per unique email, first wins
Step 5 (csv-cleaner): trim + drop empty rows + repair ragged rows

Consolidating into Salesforce: when the same contact exists in more than one CRM, keep the Salesforce row. There is no keep-last toggle in the runner, so this is controlled entirely by input order.

Step 1 input order (decides the winner at Step 4):
  1. salesforce.csv   <- preferred, listed FIRST
  2. hubspot.csv
  3. pipedrive.csv

Step 4 (csv-deduplicator, key = Email):
  always keeps the FIRST occurrence -> the Salesforce row
  case-insensitive, so casing differences never cause a miss

Result: Salesforce data wins; HubSpot/Pipedrive only contribute
emails Salesforce did not already have.

Pipedrive stores Name as one combined field. Split it BEFORE Step 1 so the merged file already has First/Last columns to rename.

Pipedrive raw:
  Person ID,Name,Email
  12345,John Smith,john@example.com
  12346,Maria Garcia Lopez,maria@example.com

Pre-step: csv-column-value-splitter on Name (space) ->
  Person ID,First Name,Last Name,Email
  12345,John,Smith,john@example.com
  12346,Maria,Garcia Lopez,maria@example.com   (greedy: rest -> Last Name)

Then drop the split Pipedrive file with the others onto Step 1.
Multi-word FIRST names (Maria Jose) still misplit -> fix manually.

The same person stored with three different email casings still dedupes to one row — because Step 4 is always case-insensitive, you would get this even without Step 3.

Salesforce: Email = John.Smith@Gmail.com
HubSpot:    Email = john.smith@gmail.com
Pipedrive:  Email = JOHN.SMITH@GMAIL.COM

After Step 1+2: 3 rows, 3 casings
After Step 3:   3 rows, all john.smith@gmail.com (output tidy)
After Step 4:   1 row (deduper lower-cases the KEY internally,
                so it would collapse these even if Step 3 were removed)

The deduplicator passes blank-key rows straight through instead of merging them, so partial CRM records aren't silently dropped — but they also aren't deduped against each other.

Merged rows:
  First Name,Email
  John,john@example.com
  John,john@example.com   <- duplicate, collapses
  Jane,                   <- blank email, PASSES THROUGH
  Jane,                   <- blank email, ALSO passes through (NOT merged)

After Step 4:
  John,john@example.com   (1 of 2 kept)
  Jane,                   (both blank-email Janes survive)

If blank-email rows must be reviewed, run csv-find-replace or a
column filter after Step 5 to isolate rows with an empty Email.

csv-merger throws CSV Merge requires at least one file input when no file reaches its csv[] port — the run aborts on the first step. This happens if you forget to stage your exports or wired the entry node downstream of an empty branch. Drop your CRM CSVs onto the merger node, then re-run.

If you switch Step 1 to strict, the merger rejects files whose columns don't line up — which is exactly the situation here, since each CRM exports different headers. Keep mode: union so every column is preserved with blanks; strict is for already-identical schemas (see the GA4 / Search Console merge workflow where strict is appropriate).

csv-deduplicator's match field expects a column NAME held as a tag/array. The blueprint stores Email as a bare string, so confirm Email is selected in the deduplicator's 'Match columns' field on the canvas before running — otherwise the key resolution can fall back to the first column and dedupe on the wrong field. Re-pick Email and the runtime keys correctly.

The orchestrator deduplicator always keeps the FIRST occurrence; its strategy config is not read at run time. To make the newest record win, sort the merged file by Last Activity Date descending with `csv-sorter` BEFORE Step 4, so the freshest row is the first one the deduper sees.

Rows with a blank Email key are passed through unchanged — they are neither dropped nor deduped against each other. After the run, isolate them with a column filter on an empty Email and resolve manually (assign emails, or dedupe on a fallback key like Phone in a second pass).

The blueprint sets normalizeSmartQuotes and normalizeHiddenWhitespace on csv-cleaner, but the orchestrator's cleaner handler only reads trimWhitespace (it always trims, removes empty rows, and repairs rows). If you specifically need curly-quote folding or NBSP normalisation, run a dedicated `csv-find-replace` pass after Step 5 to target those characters.

Pipedrive stores names in a single Name field while Salesforce and HubSpot use separate First/Last columns. Split it with `csv-column-value-splitter` BEFORE dropping it onto Step 1 — first word is First Name, the rest is Last Name. Greedy split handles Garcia Lopez correctly; compound first names (Maria Jose) misplit and need a manual fix.

Salesforce uses 15-char case-sensitive IDs internally; the API/UI often append a 3-char checksum to make 18-char IDs. Both refer to the same record. If you dedupe on Account ID instead of Email, normalise length first with `csv-find-replace` (truncate to 15 or pad to 18) — otherwise the same record appears as two rows.

HubSpot zips CSV exports over 2 MB. Extract the ZIP before dropping it onto Step 1 — the workflow does not auto-unzip. The extracted CSV merges cleanly with the others through union mode.

Salesforce custom fields (Industry__c) and HubSpot custom properties don't share a namespace. Union mode keeps them all — rows from other CRMs get empty cells in those columns. Decide per field whether to keep it (add the alias to Step 2's map or leave it as-is) or drop it with `csv-column-value-splitter`'s sibling column-remover before importing to your target CRM.