Im Feld Manager stand die Personalnummer der vorgesetzten Person -- also der
Wert, der in derselben Zeile als Local System ID gefuehrt wird. Cornerstone
verknuepft aber ueber die User ID. Der Verweis lief damit entweder ins Leere
oder, schlimmer, auf jemand anderen, dessen User ID zufaellig aussieht wie
eine Personalnummer.
Leer, wenn die vorgesetzte Person selbst keine Cornerstone-ID traegt. Aus
demselben Grund wie bei User ID und Username: eine Kennung der falschen Art
ist schlimmer als keine, weil niemand ihr ansieht, dass sie falsch ist. Bei
435 von 785 Personen fehlt die Kennung noch -- das ist eine Luecke in der
Zuordnungstabelle, kein Fehler im Export.
Die Abbildung im Kontext heisst entsprechend managerKennung und traegt jetzt
Zeichenketten, damit der Typ selbst keine Nummer mehr zulaesst.
Reworked against the load spec and the 27.09. test file. The values I
had guessed were the German display names, which is the first entry on
the list of errors from earlier loads: Cornerstone answers "ungültiger
Wert" and rejects the whole row, followed by "Alle abhängigen Felder
müssen gültig sein" as a follow-on.
Status Aktiv/Inaktiv -> Active/Inactive
Employment Status Arbeitend/... -> Working/On Leave/Terminated
User Type Mitarbeiter -> Employee
Time Zone CET -> empty
The time zone is the second entry on that list: only a portal time zone
id is valid, an abbreviation gives "Zeitzonencode nicht eindeutig".
Empty means the portal or the OU decides.
Division ID is the GUID from the test file, not a name. Termination
fields and Leave Reason are filled only when the employment status
carries them -- a reason without a termination is an invalid state for
the load, not extra information.
Four fields now stay empty on purpose, because filling them would mean
inventing an identifier that belongs to the target system: Location ID
(locations has id/name/country and no Cornerstone id), Position ID (our
S-0001 is not a Cornerstone position), Months of Service (Cornerstone
derives it) and Rehired Employee (the value for "yes" is unconfirmed,
and an unconfirmed value costs the whole row). Retention Rules and
Organisationsstufe stay empty because the load ignores them.
Header and row now match the test file byte for byte, except User ID
(no TEST- prefix outside a test load), Required Training Approvals and
Location ID.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The 57 columns of the Cornerstone user import, in the given order and
spelling, offered as CSV and Excel under the Honestly report. Same
people and same filters as the full export.
Two things about the file itself, both of which would have failed
quietly. Cornerstone reads comma-separated, while toCsv writes the
semicolons and the BOM that German Excel wants -- the BOM would have
hidden inside the name of the first column, so "User ID" would have
matched nothing in the mapping. toCsv now takes the form as an argument
and keeps its old defaults; the Cornerstone form lives next to the
columns so a test can hold both. Quoting follows the delimiter now,
otherwise a comma in an address would split the row.
Dates go out day-first, matching the import setting. An ISO date is
read as a different, equally valid date and nobody notices.
Gender maps to Cornerstone's own values; anything unexpected becomes
"not specified" rather than empty, because an invalid value makes
Cornerstone reject the whole row, not just the field.
Status and Employment Status come from separate rules: somebody on
Karenz has a working account and is not working, and filling both from
one value gets one of the two wrong.
Four fields carry visible placeholders because the leading systems do
not supply their identifiers yet: Division ID, Doxis, Interflex,
LGVplus. Home Phone and Personal Email stay empty on purpose -- what
leaves the house is the business data.
Not verified against a running Cornerstone import, and not seen in a
browser; there is no database reachable here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
