Import Multi-Value Fields Without Losing Meaning
Test append versus replace behavior, preserve unknowns, and compare actual option sets rather than flattened strings during multi-value CRM imports.
10/23/20263 min read
A company can serve both North and West without those regions becoming one new category called “North, West”. A contact can have two approved interest labels without either label replacing the other. Multi-value fields lose meaning when an export turns a set of choices into an undifferentiated string and the next importer guesses how to split it.
Define the destination's multi-value representation explicitly. Then test both the values and the update operation. Appending one option to an existing set is a different instruction from replacing the entire set, even when the incoming file looks the same.
Separate values, separators, and operations
Record the destination property type, permitted options, accepted import syntax, separator escaping, and whether the selected method appends or overwrites. Check how an empty cell behaves. It may preserve existing values, clear them, or fail validation depending on the tool and mode.
HubSpot's import guidance distinguishes multiple-checkbox append behavior from overwrite syntax. Follow the specific property documentation and configuration rather than applying a generic semicolon rule to every CRM. A delimiter convention for one field does not establish the meaning of another.
Decide whether order matters. Service regions are often a set, while ranked preferences are an ordered list. If order matters but the destination stores only an unordered set, a successful import still loses information. Choose a supported model before loading the data.
Build a fictional state-transition fixture
Suppose a synthetic account, M01, currently has the approved labels North and West. The team wants to add Central while preserving the current pair. M02 has South, but a reviewed correction should replace it with East. M03 has no confirmed region.
M01: before North plus West; operation append; incoming Central; expected North plus West plus Central.
M02: before South; operation replace; incoming East; expected East only.
M03: before unknown; incoming blank; expected unknown, not an empty set interpreted as no coverage.
M04: incoming North twice; expected one valid North option if the destination uses set semantics.
M05: incoming an unrecognized option; expected quarantine or documented property error.
M06: incoming a legitimate label containing punctuation; expected one label, not accidental splitting.
The fixture distinguishes unknown coverage from confirmed absence. If those states matter commercially, represent them separately. An empty region field should not silently mean that the account serves nowhere.
Verify by parsing the returned set
After the pilot, export the property and interpret it using the documented destination representation. Compare the resulting option set with the expected set for each record. Do not compare raw serialized strings if the application legitimately changes option order.
For M01, finding Central proves only that the new value arrived. Check that North and West remain. For M02, finding East is insufficient unless South is gone. These paired assertions catch the common mistake of using append when replacement was intended, or replacement when append was intended.
Check any reports, filters, or routing logic that consume the field. A pasted display string might look correct in a record view but fail a filter expecting actual enumerated options. Confirm that the intended option is selected through the supported property model.
Handle uncertain conversions without flattening them
Imagine an old tracker stores “North/West” with no documentation. The slash could mean both regions, a combined territory, or an undecided choice. Do not split it simply because the destination accepts multiple values. Ask the data owner to establish the original meaning, and quarantine the ambiguous row meanwhile.
If a source contains free-form phrases rather than a controlled vocabulary, create a reviewed mapping and retain the original reference where necessary. Never create hundreds of new picklist options from arbitrary notes merely to avoid errors.
Finish with an operation-aware mapping: source value, intended set, append or replace instruction, expected prior state, approved destination syntax, and exception decision. Version it alongside the importer settings. A reliable multi-value import preserves the relationships among values and the intended change, not just the words that happened to appear in the source cell.
If you are also choosing research access, ask about Leads Automation’s independent Sales Navigator offer at US$15/month on WhatsApp. Confirm the available plan, eligibility and terms before ordering; this is not LinkedIn’s official price or an assurance that any particular integration is included.