HubSpot + ServiceTitan integration

ServiceTitan customers are rarely small. By the time a contractor is running it properly there are several trades under one roof, a call centre, a membership base worth defending and often two or three acquired brands that still answer their own phones. The platform runs all of it well. What it does not do is hold the relationship above the job.

That gap shows up first in marketing. Marketing Pro builds audiences from ServiceTitan’s own data, and for a single brand selling to homeowners that is frequently the whole answer. The limit appears when the same person is a residential customer of one brand and the facilities contact for a commercial account at another, or when a private equity owner wants one view of every brand’s pipeline. ServiceTitan is deliberately tenant shaped, and a group of brands is more than one tenant.

The second driver is commercial work. Residential service is transactional and ServiceTitan models it beautifully. Commercial sales are long, involve several people at one customer, and need a pipeline with stages, tasks and a forecast that a field service platform was never built to hold. Contractors end up running that half of the business on a spreadsheet, which is what usually prompts the call.

Seven objects cover almost every engagement. Customers become companies, contacts become contacts, locations become a custom object associated to the customer, estimates become deals, and jobs, invoices and memberships become custom objects associated down the chain.

The estimate is the object worth arguing about. A job is work that already exists, so it is a record to display, not a pipeline to forecast. An estimate is a thing someone is trying to win, which makes it the only ServiceTitan record that behaves like a deal. Mapping jobs to deals looks convenient in week one and produces a pipeline nobody can read by week six.

Memberships earn their place for a different reason. The expiry date is the most valuable single field in this integration, because it turns a renewal into a HubSpot workflow with a sequence behind it. Every object keys on its ServiceTitan id, held in a custom property and used as the match key for batch upserts, which HubSpot accepts a hundred records at a time.

The shape is polling with a watermark. There is no general event stream to subscribe to, so a scheduler walks each object in pages, keeps the position it reached in a durable store and resumes from there. Each run maps records, compares a content hash against the last write and skips anything unchanged, which keeps the write volume proportional to real activity rather than to the size of the tenant.

Tenant scope is explicit everywhere. Client credentials are issued per tenant, so the service holds a credential set for each brand and every write carries the tenant it belongs to. A portal allowlist that fails closed sits in front of the HubSpot client, so a misconfigured environment writes nothing at all rather than writing to the wrong portal. A reconciliation endpoint re-reads a window on demand, which is how an outage gets repaired without a manual export.

Start with the question of who owns the customer. If ServiceTitan owns it and HubSpot only needs to send campaigns, Marketing Pro is cheaper, closer to the data and better at what it does. Nothing here beats it on that job.

If HubSpot has to own the relationship, the next question is how many tenants. One tenant with one trade can be pushed through a per-task automation platform as a pilot and will work. Several brands, or a commercial pipeline with its own stages, needs somewhere to keep watermarks, tenant scope and a mapping that survives a schema change. That is a written service.

1. Scope. We list the tenants, the trades and business units in scope, and the fields each team actually uses, then decide whether Marketing Pro already covers the requirement.
2. Architect. The object model with estimates as deals, the key on every ServiceTitan id, and the watermark strategy per object.
3. Build. The reader, mapper and writer against the integration environment, which is a clone of production for customers who have one, and a HubSpot sandbox portal.
4. Validate. A full backfill with counts reconciled per object, then the duplicates and the membership edge cases found in discovery.
5. Operate. Watch the run history, replay windows after an outage, and revisit the mapping when a new brand joins the group.