INTEGRATION
Systems that agree on what happened
An integration isn’t finished when data moves. It’s finished when two systems still agree eighteen months later, and someone knows what to do when they don’t.
Most broken integrations were built without a contract. Nobody wrote down which system owns the phone number, what happens when a record is deleted on one side, or who gets told when a sync fails at 2am. So the sync fails quietly, and six months later the finance team stops trusting the CRM.
We start every integration with a written contract per system pair: fields, direction, ownership, frequency, transformation rules, conflict resolution, error handling, retry and replay, and the named human who is paged when it breaks. The code is the easy part after that.
What you get
- A record owner for every field. Conflicts are resolved by rule, not by whoever synced last.
- Failures that announce themselves. Dead-letter handling, alerting and one-click replay.
- Governor-limit-aware by design. Bulkified, async where appropriate, tested against realistic volumes.
- Middleware you can change. No logic buried where only one contractor could find it.
- Observable. Every interface has a dashboard: volumes, latency, error rate, last successful run.
- Reversible. Every integration ships with a documented way to switch it off safely.
What we integrate
ERP (SAP, NetSuite, Dynamics, Business Central) · Finance & billing (Xero, Sage, Stripe, Chargebee) · Support (Zendesk, Jira Service Management, Intercom) · Marketing (Marketing Cloud, HubSpot, Braze) · Data platforms (Snowflake, BigQuery, dbt) · Telephony & CTI · E-commerce (Shopify, BigCommerce, custom) · EHR/EMR via HL7 or FHIR · Identity (Okta, Entra ID, SSO/SCIM) · Bespoke internal systems over REST, SOAP, SFTP or message bus
Patterns we use: REST & Composite API · Bulk API 2.0 · Platform Events & Change Data Capture · Salesforce Connect / External Objects · MuleSoft · iPaaS (Workato, Boomi, Zapier for low-stakes edges) · Webhooks with signature verification · Middleware queue with retry and dead-letter
How we run it
01 Contract — Field-level mapping and ownership agreed and signed off. 02 Prototype — One end-to-end record path in a sandbox, with deliberate failure injected. 03 Build — Bulk-safe implementation, transformation, logging, alerting. 04 Volume test — Run at 3× projected peak, then measure. 05 Cutover — Backfill, parallel run, reconciliation report, switch. 06 Observe — Dashboards, alert thresholds, and a monthly interface review.
What you receive
Integration contract per system pair · Sequence and error-path diagrams · Middleware configuration in source control · Volume test results · Monitoring dashboard · Runbook for the three most likely failures · Replay procedure · Handover session
Common questions
Do we need MuleSoft?
Only sometimes. MuleSoft earns its cost when you have many systems, reusable APIs and a governance need. For three point-to-point interfaces it’s usually overkill, and we’ll say so.
Real-time or batch?
Whichever changes a decision. Real time costs more to build and more to operate; we use it where a human is waiting on the data and batch where they aren’t.
What about API limits?
We size against your org’s limits during design and build with bulk and async patterns. Limit headroom is part of the volume test report.
Can you fix an integration someone else built?
Frequently. We start by reconstructing the missing contract, which usually explains the failure on its own.
How do you handle deletions and merges?
Explicitly, in the contract. Soft-delete, tombstone events and merge-survivorship rules are decided before build, because they’re the most common source of silent divergence.
Bring us your least trusted data flow
We’ll map the contract that was never written and tell you what it would take to make the two systems agree.
Start here
Tell us what your CRM is getting wrong
Thirty minutes with a consultant, not a salesperson. You’ll leave the call with at least one thing you can fix yourself — whether or not you hire us.
Prefer email? hello@zeptglobal.com · Typical reply time: one business day