The CSV That Opened as Garbage
Migration day started well. A Guangzhou operator running apparel across three stores — about 160 SKUs, split between the US and Southeast Asia — exported the creator list from the old tool at 9am. By lunch the file was open in Excel and every non-Latin handle had turned into a row of broken characters. The Thai and Vietnamese names were unreadable.
He fixed the encoding and imported anyway. The handles came back. Then he looked for the rest of it and found nothing.
Gone: two years of conversation notes, the tier labels marking who was worth samples, every sample shipment record, and the one-off 22 percent rate he had negotiated with a creator in Bangkok who had been his best performer for eight months.
The handles survived. The memory did not. And the handles alone are worth very little, because anybody can rebuild a handle list in an afternoon.
What hurt most was the tier labels. He had spent two years marking which creators posted without being chased, which ones needed a sample before they would film, and which ones quietly returned everything. That judgement was the asset. It lived in a dropdown menu nobody thought to export.
What This Covers and What It Does Not
This guide covers the four things that usually break during a creator list migration, a pre-export checklist, field mapping rules, and five verification passes to run afterward.
It does not cover contract migration, payout history, or content rights, which involve legal review rather than data work.
The advice assumes you are moving from one tool to another while continuing to operate. If you are starting fresh with no history, skip to the last section.
One hard limit up front. A migration can only move what the old tool actually stored. If your team kept its real notes in chat, no export will recover them, and this article cannot help you there.
The Four Things That Break in a Move
Every migration I have watched lose something lost it in one of these four places.
| What gets lost | Why it breaks | How to check |
|---|---|---|
| Contact history | Stored as notes, not fields, so exports skip it | Open ten creators, count messages per record |
| Tier and tag labels | Old taxonomy has no match in the new tool | Compare tier counts before and after |
| Sample records | Linked to orders, not to creators | Check that shipments still show a post status |
| Commission agreements | Negotiated rates live in messages | Search for “rate” and “percent” in old notes |
The pattern is that three of these four are relationships between records. Handles, names, and follower counts export cleanly because they are plain fields. Anything that connects one record to another is where the move fails.
Why It Matters More in Apparel
Apparel programs carry more history per creator than most categories, because sizing, colourways, and restocks all generate follow-up. A handle plus a follower count tells you nothing about whether that creator already tried your autumn line and found the sizing off.
Before You Export: A Pre-Flight List
Most migration damage happens in the first ten minutes. Slow down there.
The instinct is to export everything, import it, then clean up what looks wrong. That order guarantees you will not know what went wrong, because you have nothing to compare against. Export a sample first, verify it, and only then run the full file.
Also decide who owns the move. Migrations fail more often from two people editing the same export than from any technical fault.
Pin Your Field Inventory
Open the old tool and write down every field attached to a creator record. Not the ones you think matter. All of them. Include the ones you never look at, because those are the ones nobody remembers to map.
Do this before you click export, not after. Once the file is on your desktop, the inventory becomes guesswork, and guesswork is exactly how the tier labels went missing.
Then mark each as move, derive, or drop. Drop is a valid answer, but make it a decision rather than an accident.
Run a Ten-Row Sample First
Export ten records. Include at least two with non-Latin characters, one with a sample in transit, and one with a negotiated rate. Import those ten into the new tool and check them by hand.
If those ten break, all fifteen hundred will break the same way.
Check the file in a plain text editor too, not just a spreadsheet. Spreadsheets silently convert some values, especially long numeric IDs and handles that start with a symbol. What looks fine in a grid may already be altered before import. Finding that at row ten costs twenty minutes. Finding it after the full import costs a day.
The Six Checks Before the Full Export
Run these in order, and check each one against the sample file rather than the full export.
Confirm the encoding is UTF-8. Confirm dates carry a timezone or a consistent format. Confirm handles are unique and not names. Confirm the export includes notes and tags, not just profile data. Confirm sample records carry a link back to the creator. Confirm you have a raw copy you did not edit.
Keep the raw copy somewhere you will not clean it up. Edited exports cannot be re-imported when the first attempt fails.
Mapping Fields Without Guessing
Mapping is where good exports turn into bad data. It is also where the damage is least visible, because nothing looks broken at first.
Four rules cover most of it. Map on a stable key, keep the original value, never invent data, and write the mapping down where the next person will find it.
Use the Handle as the Key
Never map on display name. Names change, and two creators can share one. The handle is the only stable identifier you have across tools.
Keep the Original Beside the Mapped Value
For every field you transform, keep the original in a raw column. When something looks wrong in month three, you will want to know what the old tool said.
Do Not Invent Data
If the old tool has no first contact date, leave it empty. A guessed date is worse than a blank, because it looks reliable and it will end up in a report.
Watch Three Specific Fields
Three field types cause most mapping failures, so check them by hand on your ten-row sample.
Dates break on format and timezone. Tags break when the new tool has a different taxonomy, so map old tags to a plain text column before you try to fit them into new ones. Notes break on character limits, so check that nothing got truncated at import.
After the Move: Five Verification Passes
Do these the same day, in this order.
Pass one: counts. Compare record counts. Old and new should match exactly. A mismatch means something failed silently, not that you have fewer creators.
Pass two: handle match. Check how many handles resolved. Anything below 100 percent needs a list of the failures.
Pass three: relationship integrity. Open ten creators and confirm that notes, tags, samples, and rates are attached to the right person. Wrong attachment is the failure mode that counts will not catch.
Pass four: date sanity. Look for dates in the future, dates in 1970, and dates that all share one timestamp. Any of those means the format was misread.
Pass five: distribution comparison. Compare your tier counts, or post counts, against the old tool. If the totals match but the shape is different, your mapping reorganized something.
Log the failures rather than fixing them in place. A list of broken records tells you whether the problem is systematic, such as one field mis-mapped, or scattered, which usually means a bad source export. Those two need different fixes, and guessing wastes a day.
Spot-Check Ten, Not Two
Ten is not arbitrary. Two records will pass by luck. Ten records will surface a mapping error, a wrong attachment, and usually a truncation. Budget an hour.
Pick the ten deliberately rather than at random. Include your two highest-value creators, one whose notes are unusually long, one with a sample still in transit, and one with a negotiated rate. Those are the records where a silent failure costs the most.
Keeping History Alive After Day One
Migration is not finished on import day. It is finished when the new tool holds six weeks of history that nobody had to reconstruct.
Three habits do that. Log outreach in the tool rather than in chat. Attach sample state to the creator record the day it changes. And keep the mapping document next to the export, dated.
Then run one parallel check. For the first thirty days, keep the old tool readable and compare one number weekly. When two consecutive checks match, stop.
Be realistic about what survives, though. Free-text notes written by four different people over two years have low value even when the migration is perfect. Moving them is worth an hour. Rebuilding them is not worth a week.
Decide now what the new tool must record, then hold that line. A migration is the one moment when the team will accept a new habit, and it passes quickly. If outreach logging is optional on day one, it will be missing by day sixty.
This is the part where the destination tool matters. DAMI’s team overview and multi-store collaboration keep contact history, sample state, and agreed rates on the creator record across stores, so a second migration is less likely to be needed. DAMI’s email campaign tracking also means outreach history lands in the tool instead of somebody’s inbox.
Frequently Asked Questions
What file format survives a move best?
Plain UTF-8 CSV beats everything else. It imports into every tool, it does not carry hidden formatting, and you can read it in a text editor when something looks wrong.
Can I keep negotiated commission rates in the new tool?
Only if you recorded them as data. Search your old tool for rate notes before you export, and put each one into a dedicated field rather than a message.
How do I verify that nothing was dropped?
Run the five passes above on the same day as the import. Counts first, then handle resolution, then ten manual spot-checks.
Should I migrate everything or start clean?
Migrate relationships and history. Drop stale profile fields you never used, such as follower counts from two years ago. Old numbers presented as current cause bad decisions.
How long should I keep the old tool running?
Thirty days at minimum, with a weekly comparison. Ninety days if the old tool holds anything you could not export, which is more common than sellers expect.
Move the List, Keep the Memory
The Guangzhou operator rebuilt his tier labels in about six weeks, one creator at a time, mostly from memory. He never recovered the two years of notes, and he stopped trying once he accepted that the reconstruction was going to be more fiction than record.
The handles came back in an afternoon. He never got the two years of notes, because the old tool had only ever stored them as unstructured text and nobody had exported them before cancelling.
Do the unglamorous part first. Inventory your fields. Export ten rows. Import them, check them by hand, and only then run the full file.
For the export mechanics on the affiliate data side, the guide to exporting affiliate data covers the fields most teams forget. If you want contact history, samples, and rates held in one place so this does not happen twice, see how DAMI structures the creator record on the product page, or create your account and import ten rows to test it.
