Part of the Operations series on the Mobtag platform.

In short: Imports do not fail because CSV is “hard.” They fail because columns wander, “everyone’s template” is actually four templates, and nobody has piloted a failure mode before the big bang run.

A strong import is a ritual between operations and IT: a frozen column list, a dry run on 100 rows, then 1,000, then a quiet window for the 10k line. If your plan is “one heroic upload Friday,” you are planning a heroic weekend, not a cutover.

A sequence that works more often

  1. Name the file contract: column order, null rules, and what a duplicate serial means (error vs merge).
  2. Validate a small file in the same path the big file will use. Fix the template, not the rows by hand, when the same class of error appears twice.
  3. Report rows that fail in a file your team can fix in Excel without shame—then re-import only what failed.
  4. Freeze a release window and have a back-out: snapshot or a restore story someone has tested.
flowchart LR T[Template agreed] --> S[Small batch] S -->|Errors| F[Fix template or source] F --> S S -->|OK| M[Larger batch] M --> G[Go-live cutover]

Summary: A good import is weeks of calm alignment on the shape of the data, not a one-off upload. That is what saves the re-key and the nervous IT call at 2 a.m.