Prospects ask us this in nearly every deal, and almost nobody publishes an answer.
I understand why. Integration scope varies wildly, every partner is afraid of being held to a number, and "it depends" is technically true. It's also useless to a buyer trying to budget. So here are the ranges we gave a PE-backed manufacturer evaluating a multi-division rollout this summer, along with the variables that move each one. Take them as planning ranges, not quotes.
A native integration, using HubSpot's connector (NetSuite, for example) with standard field mapping and the workarounds that connector needs: $1K to $5K.
Custom middleware to an ERP without a native connector, Oracle Fusion in this case, with transformation logic between the two systems: $4K to $12K.
A fully custom build, where the integration carries real business logic such as quoting, approvals, or product catalog work on top of the sync: $15K and up.
Those are wide bands on purpose. What follows is what puts you at the bottom or the top of each.
Direction of sync. One-way, ERP into HubSpot, is the cheap end. Bidirectional, where HubSpot writes back to the ERP, roughly doubles the testing burden, because now you can corrupt the system of record.
Historical load versus live sync. These are two different problems that get quoted as one. Loading seven years of sales history once is a data project. Keeping tomorrow's orders in sync is an integration project. Most manufacturers need the first far more than the second, and a nightly one-way load with a few derived fields often beats real-time middleware entirely.
Lookup fields. If the fields you care about reference other tables in the ERP, the native connector may not sync them, and each one needs a workaround field and a translation workflow. Ask which of your fields are lookups before you ask for a quote.
Data quality. Clean source data sits at the bottom of the range. Data that needs transformation, deduplication, or reconciliation between systems pushes toward the top, and sometimes rules out the native connector altogether. Native connectors move data. They don't fix it.
Who maintains it. A native connector is maintained by the vendor. Custom middleware is maintained by someone, and if that someone is your two-person dev team, the honest quote includes enablement time so they can actually own it.
On one project this year we quoted enterprise middleware for an ERP the client was about to replace. The client, correctly, talked us down to a nightly SQL view at a fraction of the scope. Ranges are only as good as the discovery behind them, which is why we won't give a fixed number before we understand the process the integration serves. Anyone who will is guessing, and the guess shows up later as a change order.
If you want to pressure-test your own scope, we've published a companion piece on what a Salesforce to HubSpot migration actually costs, with a calculator. The two projects often travel together.
Where does your integration land in these ranges, and what pushed it there? I'd like to know if the bands hold outside our own client base.