HubSpot Marketing and Advertising Integrations
Marketing integrations are bought to answer one question: which spend produced revenue. Most of them answer a different and easier one, which is which spend produced a form submission.
The distance between those two answers is where the category's difficulty sits. A form submission is an event the advertising platform can see. Revenue is an event months later, on a record the platform has no access to, frequently under a different name at a different company.
15 systems in this category, 2 of them documented in a published case study.
Attribution is a modelling choice, not a measurement
Every attribution number is the output of a rule about how credit is divided, and the rule is chosen rather than discovered. First touch, last touch and even distribution across touches produce different answers from identical data, and none of them is wrong.
What matters is that the rule is stated and held constant, because a number compared against itself under a changed rule is not a comparison.
The second and larger problem is coverage. Attribution can only apportion credit across touches it recorded. Offline conversations, referrals and anything that happened before the tracking cookie was set are invisible, so the model apportions the whole of revenue across the fraction of reality it can see.
The same journey under three attribution models, producing three answers
What the model can see against what actually influenced the decision
Systems in this category
Engagement signals joined to the account, deduplicated on arrival.
Enrichment written to named properties rather than over whatever it finds.
Cadence membership and outcomes reflected on the contact.
Sequence state kept in one place instead of two competing ones.
Programme membership and scoring migrated with the definitions intact.
Audience and subscription state reconciled with HubSpot's own.
Commerce behaviour joined to the contact without duplicating consent.
List history preserved through consolidation.
Form submissions and page activity resolved to a single contact.
Migration with redirects mapped one to one and verified after cutover.
Registrations and attendance recorded as separate facts, because they are.
Attendance duration written back, not merely registration.
Audience membership driven by CRM state, and closed revenue posted back so bidding sees revenue rather than form fills.
Offline conversions returned so bidding sees revenue rather than form fills.
Conversions API fed from closed revenue instead of page events.
Documented engagements
Each of these is a published study of one build, with the object model, the decisions taken and what the approach does not establish.
Common Questions
Frequently Asked Questions
None of them, in the sense of being true. Each is a defensible rule for dividing credit. The useful practice is to pick one that matches how the business actually buys, state it where the report is read, and change it rarely.
Because they are counting different things over different windows with different identity rules. The advertising platform counts a conversion it can attribute to a click within its own lookback window. HubSpot counts a contact it can join to a deal. Neither is broken and reconciling them exactly is not achievable.
Where the spend is meaningful, yes. Sending closed-won values back lets the platform optimise toward revenue rather than toward form fills, which changes what it buys. It also requires an identity path from the advertisement to the deal that survives months, which is the part that takes the work.
Further reading
Long-form already published on the argument this category turns on.
Related categories
A system in this category that is not listed
The directory shows range rather than limits. A scoped build is quoted from the object model on each side and the fields that have to cross between them.