Resources | Ampersand - Native product integrations: when to build and when to buy

Build vs buy for integrations: the real cost

The first version is the easy part. Here’s what building integrations actually cost once maintenance, opportunity cost, and scale enter the picture.

Talk to an engineer

3-YEAR COST, THREE CONNECTORS

Build in-house

  • Build
  • Maintenance

With Ampersand

  • Implementation

When teams face the build-or-buy decision

The first integration looks manageable: one API, one authentication flow, one data model to learn. Each connector after that brings its own authentication quirks and its own failure modes, and turns into a system your team supports for years, not a project you finish.

  • ONE API
  • One auth flow
  • One data model

Years of support

The real question isn’t whether your engineers can build another connector. It’s whether that work belongs on your roadmap, or pulls engineers off the product you’re actually selling.

Estimated connector build cost by API complexity

Using a $200,000 annual engineering cost as the baseline, here’s what one connector typically takes to build, from auth through customer setup.

BASELINE $200K / eng-year

  • Tier 1 Standard API scope

    Typical scope
    Standard OAuth, a limited set of objects, predictable pagination, and straightforward error handling
    Example providers
    HubSpot or Slack for a narrow use case
    Build estimate
    1 to 1.5 engineer-months
    Est. engineering cost
    $17,000 to $25,000

  • Tier 2 Provider-specific complexity

    Typical scope
    Endpoint-specific rate limits, varied pagination, additional permission cases, and more extensive error handling
    Example providers
    Zendesk, Jira
    Build estimate
    2 to 3 engineer-months
    Est. engineering cost
    $33,000 to $50,000

  • Tier 3 Highly configurable platform

    Typical scope
    Customer-specific objects and fields, API quotas, bulk data operations, and multiple API surfaces
    Example providers
    Salesforce, NetSuite
    Build estimate
    4 to 6+ engineer-months
    Est. engineering cost
    $67,000 to $100,000+

Three connectors, one from each tier, run 225K+ before maintenance starts.

Then maintenance starts

Why maintenance can cost even more than the initial build

The build is roughly 30% of what a connector costs over its life. Maintenance is the other 70%. An $80,000 build implies about $187,000 in maintenance over three years, close to $62,000 a year, for a three-year total near $267,000.

Lifetime cost of one connector · example

3 years

Build · 30%

$80,000

Maintenance · 70%

$187,000 close to $62,000 a year

Three-year total ~$267,000

Systems of record: no two customers configure them the same

  • Systems of record like Salesforce and NetSuite are particularly expensive because no two customers configure them the same.

  • Systems of record have been around for many years and often accumulate custom fields and objects, different permissions, and different workflows.

  • There’s no generic integration that works for every account. Every new customer’s system of record integration will require your engineers’ time to set it up and customize it, and that work repeats with every account you add.

Once it’s live, maintenance breaks down into four recurring jobs:

Three connectors, three years

Opportunity cost and the breakeven point

Buying wins for any multi-connector roadmap. Building only pays off with a single, narrowly-scoped connector, or when the integration itself is the product.

Build in-house

On Ampersand

Same three connectors

Build

When building integrations in-house makes sense

  • One stable connector, with no more requests on the roadmap.
  • You are focusing your product on only one ecosystem.
  • No vendor passes your security review, whether that’s an air-gapped environment or a data residency requirement.

Buy

When buying integration
infrastructure is the right call

  • Customers need deep, configurable integrations with one or more complex systems of record.
  • A deal is blocked on an integration, and the customer can't wait for you to build it from scratch.
  • A team with a full roadmap shouldn't have integration work competing for its time:

    Every sprint spent on integration work is a sprint not spent on your core product.

// MIT LICENSE · GO

The middle path: build on open source, buy the infrastructure

Ampersand Open Connectors is an MIT-licensed Go library you can inspect and extend yourself. The managed layer on top handles auth, sync orchestration, quotas, retries, and diagnostics, so your team keeps control of integration behavior without owning the infrastructure underneath it.

Build vs. buy decision framework for
SaaS integrations

  • Security and deployment

    Build in-house when
    A mandatory policy rules out every available vendor deployment model
    Buy integration infrastructure when
    A provider passes the security review and supports the required compliance and deployment controls

  • Product differentiation

    Build in-house when
    The connector’s depth directly influences why customers choose the product
    Buy integration infrastructure when
    Customers expect the integrations, but choose the product for its workflow, automation, or insights

  • Connector roadmap

    Build in-house when
    One connector has stable scope, with no additional providers requested or planned
    Buy integration infrastructure when
    Multiple connectors are committed or customer requests continue expanding the roadmap

  • API complexity

    Build in-house when
    The integration covers a narrow set of stable objects and permissions
    Buy integration infrastructure when
    Customer-specific schemas, custom fields, provider quotas, or multiple API surfaces increase the scope

  • Customer scale

    Build in-house when
    A small number of installations makes customer-specific support manageable
    Buy integration infrastructure when
    Dozens or hundreds of installations introduce separate credentials, permissions, and configurations

  • Data delivery

    Build in-house when
    Scheduled sync and basic recovery meet the product requirements
    Buy integration infrastructure when
    Event-driven updates, historical backfills, retries, and reconciliation must work across customers

  • Engineering capacity

    Build in-house when
    A dedicated team can own connector development, maintenance, and on-call support
    Buy integration infrastructure when
    Integration work competes with core product development and repeatedly escalates to engineers

  • Code ownership

    Build in-house when
    Company policy requires full ownership of every connector and runtime component
    Buy integration infrastructure when
    Open or extendable connector code provides enough control while managed infrastructure handles production operations

  • Three-year cost

    Build in-house when
    In-house development and maintenance cost less under the company’s own estimates
    Buy integration infrastructure when
    Platform implementation, fees, usage, and remaining internal maintenance cost less over the same period

FAQs

Build vs. buy for
SaaS integrations

Talk to an engineer
  • How long does it take to build a Salesforce integration?

    Plan for four to six-plus engineer-months for a customer-facing integration across multiple organizations. Custom objects, API quotas, backfills, and event subscriptions push that estimate higher.

  • How much does an integration cost to maintain per year?

    Under the 30/70 model, annual maintenance runs $13K to $20K for a Tier 1 connector, $26K to $39K for Tier 2, and $52K to $78K or more for Tier 3. The exact number depends on customer count and how often the provider changes its API.

  • Is open source a middle option between building and buying integrations?

    Yes. Open-source connector code gives your engineers control over provider-specific logic, and managed infrastructure handles authentication, sync orchestration, retries, and customer-level monitoring on top of it.

  • When does build vs buy break even for SaaS integrations?

    Breakeven is the point where platform fees and the maintenance you still own match what three years of in-house work would have cost. A single, narrow connector keeps the two paths close. Add a second or third connector and buying pulls ahead.

  • Do AI coding agents change the build-vs-buy math?

    They speed up scaffolding, not the part that costs the most: token refresh, rate limits, schema drift, and the debugging that shows up after launch.

Start building
deep integrations today

Talk to an engineer Start building