Your nightly sync didn't update the record. It quietly erased half of it.

An ops agent fills in a contact's 'phone' and 'lifecyclestage' by hand in HubSpot. The next sync runs, and both come back blank. Nobody touched them upstream. The integration did.

The cause is in how the update payload gets built. The sync maps the full source record onto the HubSpot object and sends every field, including the ones that are empty at the source. HubSpot's PATCH only changes the properties you include, and a property is cleared by sending an empty string. So "phone": "" doesn't mean "leave it," it means "wipe it." Every blank source field overwrites whatever a human typed on the HubSpot side. HubSpot is doing exactly what its docs say. The mapping is the bug.

The fix is to send only the fields you actually have a value for and intend to own. Strip any key whose source value is null, undefined, or "" before the request goes out. If a blank in the source genuinely means "clear this," make that an explicit, per-field decision on an allowlist, not a side effect of how your mapping serializes empties.

The principle: an update should carry only the fields it means to change. "I have no value for this" and "set this to empty" are two different operations, and a sync that treats them as one loses data without ever throwing an error. Most silent data loss isn't a failed write. It's a clean write of the wrong thing.