What a manual export involves
Moving Holded data into HubSpot by hand usually looks like this, every time:
- Export contacts and invoices from Holded to Excel or CSV.
- Clean the file: remove suppliers, fix tax IDs, split company and contact persons.
- Map each column to a HubSpot property in the import tool.
- Import, then look for duplicates and rows that failed.
- Repeat next week, because payments and new invoices have changed everything.
For a migration you do once, that's reasonable. As a routine, it's a part-time job that is always one step behind.
The four risks of doing it by hand
- Duplicates. Each re-import has to match existing records. If the matching key is missing or written differently, HubSpot ends up with two versions of the same customer.
- Stale payment status. An invoice imported as pending stays pending in HubSpot after the customer pays, so sales chase invoices that are already settled.
- Typos in tax IDs. A tax ID typed with a space or a missing letter breaks matching later and carries the error into every report.
- No history. When a figure looks wrong, there's no way to tell which import changed it or what it was before.
When doing it by hand is fine
Manual export is a perfectly good choice for a one-off migration of historical data, for a business that issues a handful of invoices a month, or for an analysis you'll do once in a spreadsheet. If nobody in sales or customer success looks at payment status in HubSpot, the risks above matter less.
Starting from existing data safely: “Link only”
Most companies that move to a sync already have customers on both sides, often from earlier imports. Holded by Vexco handles that in two steps of its setup wizard:
- Step 5, “Analyze duplicates”: compares both systems without writing anything and tells you how many records will be linked, how many would be created on each side and how many need a human decision. See the duplicates step.
- Step 6, initial load: “Link only” links what exists on both sides and creates or changes nothing; “Link and fill gaps” also fills empty fields without overwriting; “Full load” creates what's missing. See the initial load step.
You can start with a sample of 5 to 50 records per flow, or simulate up to 25 records per flow without writing anything. The wizard estimates the Holded API calls before you start; Holded's API allows 100 requests per minute per key, according to Holded's help centre, and loads never touch the reserve kept for day-to-day sync.
From spreadsheet to sync in practice
Install the app, connect Holded with an API key and follow the wizard. Activate in read-only mode first: the app only copies from Holded to HubSpot, so you can check the data for a few days and then turn on both directions from the Status tab. The setup guide covers each screen, and our article on avoiding duplicate customers explains how matching works. If your data needs a clean-up before you start, Vexco's HubSpot implementation team can help.
Frequently asked questions
Can I keep HubSpot up to date with CSV exports from Holded?
Yes for a one-off, not really for ongoing use. Each export is a snapshot: payment status and balances stop being accurate as soon as a customer pays, and every re-import risks duplicates. A sync updates them as changes happen in Holded.
We already imported Holded customers into HubSpot by hand. Is that a problem?
No. In the initial load, “Link only” links the records that already exist on both sides, by tax ID or email, without creating or changing anything, and doubtful matches go to the review inbox. Step 5, “Analyze duplicates”, tells you beforehand how many will be linked.
Will the app duplicate records we imported earlier?
No. Before creating anything it looks for an existing link, then the tax ID, then the email; if the match is doubtful, such as the same email with a different tax ID, it does nothing and sends the case to the review inbox.
How much Holded API quota does an initial load use?
It depends on your volume. The wizard estimates the Holded API calls before you start and compares them with your monthly quota; loads never touch the day-to-day reserve (20% by default) and wait if the quota runs out, then resume.