I'm implementing it · 2 min read
What to decide before you configure Data Hub, and in what order
The decision to invest in Data Hub is made, or nearly, and you need the order to decide things in before anyone touches configuration.
If you have five minutes
- Four decisions sit ahead of any screen in Data Studio, any duplicate rule, or any warehouse sync.
- Each has an owner and a cost for deciding it late, in the operators' own words where the record has them.
Decide before you configure
Four decisions sit ahead of any screen in Data Studio, any duplicate rule, or any warehouse sync. Each has an owner and a cost for deciding it late, in the operators' own words where the record has them.
| Decision | Suggested owner | What it costs to get it late |
|---|---|---|
| Which criteria define a duplicate | Whoever owns the object's data quality | Rules get built ad hoc during a merge, decided under pressure instead of in advance |
| Which field wins when two records disagree, and whether "winning" is explained to the team | The data owner, with whoever trains the team | "A record is only winning until you need a value from the other one to override it" — Casey Hawkins. Train on that definition or the label misleads on day one |
| Where transformation happens, inside Data Studio or upstream in the warehouse | Whoever owns the warehouse relationship | Credits are charged on transformation inside Data Studio, not on the sync itself, so this decision is also a cost decision |
| What "verified" means for a data-out job | Whoever owns the downstream warehouse or BI consumer | Object-level selection, scheduling, and sync health exist so this does not have to be assumed; leaving it unset returns you to the old beta's failure — not knowing whether last night's export ran |
The order the design goes in
- Decide the duplicate-rule criteria first. Create Custom Rules to Manage Duplicate Records (public beta) is what flags a duplicate before Merge Rules ever gets asked which one wins.
- Decide the merge criteria and the wording for "winning" second. Merge Rules | Winning Record Rules (public beta) configured without agreement on that wording becomes the record-management incident nobody wanted.
- Decide where transformation happens before turning on a warehouse-to-CRM sync. Credit Consumption Updates for Data Warehouse CRM Syncs via Data Studio (live) is what makes that decision a cost decision, not just an architecture one.
- Configure Amazon S3 Data Out (public beta) last, with object-level selection and a real schedule set deliberately, not left on a default. The value of this release is specifically that it can be watched and trusted; skipping that setup wastes the reason it shipped.
What the import fix everyone eventually builds actually costs
"You find out an import will not finish only after you are most of the way through it, which is precisely why the workarounds exist." — Chris Carolan, on Smart Duplicate Handling in Multi-Object Imports (live), first covered August 28, 2026
The sequence he describes is the one every admin eventually reaches on their own: import to get record IDs, export them back out so you have them, then use those IDs to reassociate everything across objects. The cost is not building that sequence. It is that you only learn an import will fail once you are most of the way through it — so the sequence has to exist before the failure does, not after.
Full take — Smart Duplicate Handling in Multi-Object Imports.
Where to next
Where this comes from
- Smart Duplicate Handling in Multi-Object Imports, the show's own take. Chris Carolan on the record; live, first covered August 28, 2026
- Merge Rules | Winning Record Rules, the show's own take. Casey Hawkins on the record; the source of the wording of "winning" in the table above
- The HubSpot Updates records filed under Data Hub. the duplicate rules, the merge rules, the credit change and Amazon S3 Data Out — status and first-covered date read from the record, not from a completed build
What we do not know yet
- Whether the record-ID sequence above is still needed for every multi-object import scenario now that Smart Duplicate Handling has shipped, or only for a subset of them, has not been verified against a live portal for this draft. It draws on the take, not a fresh test.
- No verified dollar-credit figure exists in the material behind this door for what a warehouse sync cost before this change compared to after. Only the qualitative rule — transformation, not movement, is what is charged — is confirmed.
- Whether HubSpot's Data Studio write path (the file ingestion API and the data sources API, both marked live) has been exercised end to end against a real portal by anyone drawing on this door is not confirmed. The status named here comes from the record, not from a completed build.