Step 4: Migrating to Plain — import or fresh start?
Maybe you're here because you're moving from working out of an email inbox, or because you're tired of managing multiple channels in Slack — either way, you'll want to make a decision on whether to import historical data into Plain, or if it's better to work out of a clean space.
What good looks like
Ideally, by the end of this article, you’ll know whether you do, or don't, need historical data in Plain. It's best to think about this as a distinct step from new thread creation + ingestion — it's important to come to a decision, and you can make it while you connect channels and sync accounts. Run the backfill once those accounts exist (Step 3), so imported threads can pick up an account when they're created.
What to consider
There's a bit of nuance when it comes to making a decision here, but, ultimately, what we've found in 90% of use cases is that there really isn't a good reason not to import your historical threads (tickets in other platforms), especially when we have native importers for most major support platforms (we support Zendesk, Intercom, Front, Help Scout, and Freshdesk out of the box).
These importers run in the background (so you can continue with the rest of your setup) and syncs your threads over from your previous tool — the initial backfill will generally finish over an afternoon, though larger syncs of 100k+ threads can take 24-48 hours.
How to implement
Native importer
Plain support imports from most support tools right out of the box — you can see what metadata is synced over in the table below.
Configuration is easy: just click on the Settings icon in the sidebar, then on Integrations, and scroll down to the Importers section to locate your support tool and click on the Configure button.
The clip below shows Settings, then Integrations, then the Zendesk card and Connect with Zendesk Importer, which opens the confirmation dialog.
Zendesk | Freshdesk | Intercom | Help Scout | Front | |
|---|---|---|---|---|---|
Threads | ✓ | ✓ | ✓ | ✓ | ✓ |
Messages | ✓ | ✓ | ✓ | ✓ | ✓ |
Customers | ✓ | ✓ | ✓ | ✓ | ✓ |
Internal notes | ✓ | ✓ | ✓ | ✓ | ✓ |
Attachments | ✗ | ✗ | ✓ | ✓ | ✓ |
Labels (from tags) | ✓ | ✓ | ✓ | ✓ | ✓ |
Priority | ✓ | ✓ | ✓ | not stated | ✗ (always normal) |
Status | ✓ | ✓ | ✓ | ✓ | ✓ |
Assignee | if agent exists | ✓ | ✓ | if agent exists | ✓ |
Some gotchas to be aware of:
The import is a snapshot — new messages are synced over, but any status changes, assignee changes, etc. made after the initial import won't sync over. These can be updated in bulk via the API if needed.
Most importers are an all-or-nothing import — when they’re connected, we’ll import all customers and threads from your other tool.
Threads must have a customer assigned in Plain — if a customer cannot be resolved on import, that thread is skipped. This means you must import your customers from your legacy platform (which our native importers handle automatically)
Plain has built-in deduplication, so running an import multiple times won't create duplicate threads.
Imported threads won’t fire SLA policies, autoresponders, or workflows, so you can import without worrying about any downstream repercussions.
API import
If your old tool isn't in the list, a short script does the job. And no worries — if it sounds scary to you, it shouldn't! Plain is built on its own API, so the amount of manual work you need to do here is minimal. The only prerequisites you need are:
a Plain machine user holding the
thread:importpermission. You choosethread:importwhen you add an API key for that machine user.
your workspace ID
data from your previous support tool (this can either be an API key with the proper permissions, a .csv export, etc.)
If you have someone on your team comfortable writing GraphQL, then the abbreviated shape of the call can be seen below (the full walkthrough is in Importing threads). Otherwise, our recommendation is to use an LLM to build out and iterate on a script for you. Just give it the API key for your machine user in Plain, access to the original source data, and pass in our GraphQL schema here — from there, you can chat with it about things like scope, mappings, etc.
mutation { importThread(input: { customerIdentifier: { emailAddress: "ola@acme.com" } externalId: "crisp-48213" title: "Webhook retries not firing" status: DONE createdAt: "2025-03-02T09:14:00Z" }) { thread { id } } } mutation { importThreadMessages(input: { threadId: "th_01J..." messages: [{ externalId: "crisp-48213-1" type: INBOUND text: "Retries stopped overnight." createdAt: "2025-03-02T09:14:00Z" emailMessageId: "<a1b2@mail.acme.com>" }] }) { thread { id } } }
Some gotchas to be aware of:
importThreadcreates the thread as a backfill automatically, which prevents automations like auto-responders, SLAs, and workflows from running on import.importThreadMessagesbatches up to 25 messages per callemailMessageIdis the attribute in thethreadMessagethat keeps the conversation threaded when a customer replies to an email after it's been synced to Plain. You can pull this value from themessageIdattribute in the header of the original email.
How to verify if this step is working/complete
Ask your LLM to build a dry run mode into the script it creates — this will allow you to test the script without having to run it across all your threads.
Ask your LLM to build in a way to scope the import to "x" number of threads — this way you can test the import against your workspace. Plain has de-duplication built in, so there's no risk of creating duplicates provided the external IDs for each thread remain the same across subsequent runs.
If you're hesitant about testing against your production workspace, please reach out to Plain support for getting you set up with a Sandbox workspace to test in.
Troubleshooting/FAQ
Nothing has appeared in my workspace yet.
The first import is often quite long as we need to backfill all of the historical data in your previous tool. Imports are also done in bulk — until the run has completed, you won't see any threads in your workspace. There isn't a dedicated progress bar but you're more than welcome to reach out to Plain's support team to get a general ETA.
When do I turn off the old tool and turn on Plain?
TL;DR turn Plain on first, and turn off your old tool last (there will be a period of time where both are "on"). Here's the sequence that works for most small teams:
Kick off the import. Accounts from Step 3 are already in place, so imported threads can attach to them. Plain is not receiving live traffic yet — it's just backfilling.
Confirm the channels from Step 1 are connected, and don't redirect anything yet. You already connected your support address, Slack, and chat and sent yourself a test thread in each. Your team is still answering out of the old tool. This check is that those tests worked, and that a test thread has an account on it, before you flip live traffic.
Pick a quiet window, then flip every channel at once. Redirect the forwarding on your support address, swap the old tool's Slack integration for Plain's, switch the chat widget, etc. We'd recommend against a slow roll here (email this week, Slack next week, etc.) — it feels safer, but it often results in a scenario where nobody is quite sure of what tool to work in. Better to cut over towards the end of the week, or whenever your volume dips — you can effectively mitigate net new threads coming into your own tool if timed correctly.
Re-run the import right after you flip, and pick up in-flight threads in Plain. New threads that arrived in the old tool since the first import come across with their message history, status, and assignee. On threads that were already imported, new replies sync, but the status and assignee stay as they were at that first import — update those in Plain, or in bulk via the API. Native importers sync new messages hourly by default. They do not keep status and assignee in sync after the first import. If you're running a custom script, run it once more after you've cut your support channels over to Plain. At that point, you're safe to turn "off" your old tool. Your team can carry on responding in Plain. Done right, you're looking at 1-2 days of overlap max vs. the 1-2 weeks you would expect with a phased cutover.