What actually breaks HubSpot contact imports (notes from building a preflight checker)

I spent the last few months encoding HubSpot's contact import rules into a checker, and the list of things that actually go wrong turned out shorter and dumber than I expected. Sharing the categories in case it saves someone a re-import.

What keeps showing up:

1. Missing identity field. No email and no record ID means nothing to dedupe against.

2. Mixed date formats inside a single column. One column with 2026/03/11 sitting next to 11-03-2026.

3. Lifecycle stage and lead status values that do not match the portal. These are enum fields, so a near-miss is not a near-miss.

4. Duplicate emails inside the file itself, before it ever reaches the CRM.

5. Encoding damage. Files that went through Excel and came back with mojibake in name fields.

6. Structural CSV problems. Unquoted commas, ragged row lengths, a stray BOM in the header.

7. Phone formats. Will not block anything, will quietly make dedupe and dialers worse later.

8. Header names that look obviously right to a human and match nothing in HubSpot.

The pattern I keep hitting: most of these do not fail loudly. The import goes through and the damage shows up weeks later in reporting.

Genuinely curious what I am missing, especially from anyone doing this at higher volumes or across multiple portals.