Scoping and managing a custom integration

This article is intended for SaaS companies, consultants, IT integrators and agencies that use to provide custom integration services to their customers. We outline the different steps of a typical custom integration project, so that you can staff the right people, prepare a realistic quote and estimate the time required to deliver the overall project.

Please note that the platform is often used to create reusable integration templates, that are deployed for many customers. These templates can even be tweaked (customized) for individual customers that need a little extra. This article however, covers the case where a fully custom integration is being built from scratch, based on specific requirements from your customer, and where a template is not available (yet).

Project staffing - which roles are involved

Project Manager

Main roles:

  • Manage timeline
  • Overall communication with customer
  • Getting requirements from the customer
  • Making sure the integration is delivered, documented and tested


Main roles:

  • Defines the overall flow of data
  • Defines the object types involved (e.g. Accounts, Contacts, Invoices), the relationship between these object types, the fields involved, required transformations, how data maps from the source to the destination, business logic that needs to be applied etc.
  • Uses the scoping document (intake document) to get a more detailed view of the project


  • Good understanding of data models (relational models)
  • Good understanding of the platforms involved (CRM, marketing automation etc.)
  • Good understanding of the business case (why is the integration important, what is the required outcome, how can we verify if everything works as expected etc.)

Integration engineer

Main roles:

  • Build the actual Blend(s) in
  • Testing of the Blend
  • Switching from test accounts to production


  • Good understanding of platform
  • Completed all getting started videos and basic help articles
  • Understands the different integration Patterns, for example how to process data incrementally, what Upserts are etc.
  • Has programming or scripting skills, this is needed to build comprehensive Blends (loops, variables, conditions, data structures etc.)
  • Good understanding of JSON, (REST) API's etc.
  • Good understanding of data models (understands foreign keys, primary keys etc.)
  • Knows what epoch timestamps are, understands different date/time formats, understands challenges of timezones etc.

How to estimate the overall effort, how to make a quote

Time involved in the project

Example (actuals can vary heavily depending on complexity of the project):

  • Project management (5 technical calls, follow up etc.): 2 to 10 hours
  • Analysis: 1 to 5 hours
  • Building the Blend(s): 20 to 40 hours
  • Testing: 2 to 10 hours
  • Documentation: 1 to 4 hours
  • Finetuning/small changes after production: 0 to 20 hours
  • Ongoing maintenance, monitoring: 1 hour/month

Pricing (Quote)

Make sure to budget following aspects:

  • license for one end-customer (recurring)
  • New connectors needed ? One time setup cost needed
  • Your services based on above total amount of man hours (e.g. 30 - 80 hours)

The different phases of the project


You can use the example scoping document (intake document) to make sure you gather all the required information from the customer.

Building the integration

Use the Blend editor to build and test the integration.

Testing & QA

Make sure to build and test the Blend using test accounts on the connected platforms.

Make sure to use the Integration Checklist.

You can build a Compare Blend to compare e.g. the data in source and destination to make sure all data is synced correctly.

Bring into Production

Switch the connected platforms (datasources) from test to production.


Make sure to document your integration, especially the why's, the overal business processes involved etc.


You will receive emails if Blends fail, make sure to check the jobs (history) of the Blend on a regular basis, check for warnings etc.


You may need to make small changes to the Blend(s) (e.g. add a field mapping) as time goes by, because the data of the customer may change, more data is added etc. Make sure to also maintain the knowledge of the project so that you can dive back in when requested by your customer.