API Integration Cost: Simple, Medium and Complex
A simple API integration costs 1-2 engineer-weeks, a two-way sync 4-6, and a complex one 10-16. The cost drivers, iPaaS versus custom code, and upkeep.
An API integration between two business systems typically costs 1 to 2 engineer-weeks for a simple one-way connection, 4 to 6 for a two-way sync with business rules, and 10 to 16 for a complex integration with poor documentation, no sandbox or regulated data. At $1,400 to $2,200 per engineer-week, that is roughly $1,400 to $4,400, $5,600 to $13,200, and $14,000 to $35,200, plus a yearly maintenance budget for when either side changes its API. The spread comes less from the code that calls the API than from everything around it: understanding the other system's data, handling its failures, and proving the two systems agree. Integrations are a large share of the custom software development work we do at RNDSOL, and the estimates that go wrong are almost always the ones made from an API's marketing page rather than its documentation and sandbox. Below are the cost drivers, three worked examples, when an iPaaS tool such as Zapier or Make is the better buy, and what to budget after launch.
What Drives the Cost of an API Integration?
- Documentation quality: an API with current reference docs, examples and a changelog can be estimated in an hour; one documented by an old PDF needs days of testing to learn how it behaves
- Authentication: a static API key takes minutes; OAuth with token refresh, per-customer consent, client certificates or IP allowlisting takes days and needs secure credential storage
- Rate limits: a cap on calls per minute decides whether you can sync in real time or need batching, queues and backoff, and limits that differ between sandbox and production are a common surprise
- Webhooks: events pushed from the other system save polling, but need a public endpoint, signature checks, and handling for events that arrive twice, late or out of order
- Data mapping: matching customers, products, tax codes and statuses between two systems that model them differently is often the biggest single item, and someone from the business has to decide the rules
- Error handling and retries: what happens when one side is down, a record is rejected, or a call times out halfway, including a screen where staff can see and replay failures
- Sandbox availability: without a test environment, every test touches live data, which slows the work and pushes some testing into production
- Ongoing change: providers release new API versions and retire old ones on their own schedule, so the integration needs maintenance for as long as it runs
How Much Does a Simple API Integration Cost?
A simple integration moves one kind of record in one direction, between systems with good documentation, key-based or standard OAuth authentication, and a sandbox. Example: every paid order in an online store creates a sales invoice in a cloud accounting system, with the customer, line items and tax. The work is reading both APIs, mapping a dozen fields, handling the cases where the customer or product does not yet exist on the accounting side, adding retries and an error log, and testing with real order shapes, including refunds and discounts. That is about 1 to 2 engineer-weeks, or $1,400 to $4,400. Off-the-shelf connectors cover many cases like this one, and if one handles your two systems well, use it. Build custom when the connector cannot apply your tax or discount rules, or when you need to see and replay failures in one place alongside your other integrations.
A Medium Integration: Two-Way CRM and ERP Sync
A medium integration keeps records in step in both directions, which turns a data transfer into a set of business decisions. Example: customers and quotes are created in the CRM, accepted quotes become sales orders in the ERP, and invoice and payment status flows back to the CRM so the sales team can see who has paid. Now someone has to decide which system wins when the same customer's address is edited in both, how duplicates are matched, what happens to a quote edited after it became an order, and how records created before the integration existed are loaded. Add OAuth with token refresh, webhooks from one side and polling on the other, rate limits that force batching during the initial load, and a reconciliation report that shows records out of step. That is commonly 4 to 6 engineer-weeks, or $5,600 to $13,200, and a large part of it is mapping, conflict rules and testing with the business rather than writing API calls.
What Makes an API Integration Complex?
Complexity comes from the other side being hard to work with, or from the data being regulated or financial. Typical cases: a legacy ERP with a SOAP interface, partial documentation and no sandbox; an exchange that runs on file drops over SFTP because the old system has no API; a banking or payment connection that requires certificates, signed requests and a certification process with the provider; a government e-invoicing link where each invoice has to be validated and reported in a prescribed format; or a sync across several companies, currencies and tax regimes. These run 10 to 16 engineer-weeks, or $14,000 to $35,200, and the estimate itself carries more uncertainty, so we price them after a technical spike of a few days that tests authentication, one full round trip and the worst-documented endpoint. Our telemedicine platform case study shows what that spike is meant to catch early: the EHR vendor's production endpoints enforced rate limits the sandbox never mentioned and omitted two fields the sync depended on, and the real-time design had to be rebuilt as a queued, retry-safe pipeline.
Should You Use an iPaaS Tool or Custom Integration Code?
iPaaS tools such as Zapier, Make, n8n and Workato are the right choice when the flow is simple, both systems have a maintained connector, volumes are modest, and the person running it can see and fix failures in the tool's own interface. They are quick to set up, often in days, and need no hosting of your own. Their cost grows with volume, because most plans charge by task, operation or run, so check current pricing against your actual monthly record count, not the trial. Custom code wins when the mapping carries business logic, when volumes are high enough that usage pricing overtakes the cost of building, when the data is sensitive enough that you do not want it passing through another processor, or when the integration is part of your product rather than your back office. A sensible middle path is to start a flow in an iPaaS tool to prove it is needed, then rebuild it in code once the rules and volumes are known. If you go custom, keep each integration behind its own adapter module, so a provider's change touches one place.
What Does an Integration Cost After Launch?
Every integration has two sides that change without asking you. Providers add required fields, retire API versions, change authentication methods and adjust rate limits, usually with notice measured in months, and your own systems change too. Budget toward the top of the 15 to 25 percent of build cost a year that our software maintenance cost guide recommends for custom software, because integrations are the part most exposed to outside change. For the medium example above, that is roughly $1,400 to $3,300 a year, or one to two engineer-weeks, mostly spent on version upgrades and on investigating the occasional record that fails. Monitoring keeps the number down: alert on failed syncs and on silence, since an integration that stops sending anything has failed too, and review each provider's changelog every quarter. Without that, the cost of an unmaintained integration is usually discovered by finance at month end, when the two systems disagree.
API Integration Estimate Checklist
- Read the API's current reference documentation and changelog, not its marketing page
- Confirm a sandbox exists and find out how closely it matches production
- Identify the authentication method and who will hold the credentials
- Check rate limits against your initial data load and your daily volume
- List every record type, field and status to map, with a business owner for each rule
- Decide which system wins for each field in a two-way sync
- Specify what happens on failure, who is alerted, and how a record is replayed
- Price a short technical spike before quoting anything that touches a legacy or regulated system
- Compare iPaaS pricing at your actual volume with the cost of custom code
- Budget yearly maintenance near the top of the 15 to 25 percent band, and run the rest of the system through the cost calculator
Have a project that needs this kind of thinking applied to it?