What a custom integration costs, and how to scope one
The cost of a custom integration depends less on the number of fields and more on how many systems are involved, which way data flows, how it fails and who runs it afterwards. A one-way push of a few fields between two well-documented APIs is a small job. A two-way sync with edits on both sides, retries, alerts and an owner is a different one. Fixed price suits a scope you can write down; hourly suits discovery. Skofi's own figures are given below and labelled as ours. They are not a market benchmark.
By Saveez IrfanUpdated
In short
- Cost follows the failure surface, not the field count: systems, direction, authentication, volume, error handling, monitoring and ownership.
- Two-way sync costs far more than one-way push, because conflicts need a rule and a master for each field.
- Error handling and monitoring are usually half the work. A quote that omits them is a quote for a demo.
- Fixed price works for a scope you can write down. Hourly suits discovery or where the API behaviour is unknown.
- Ask who owns the credentials, the code and the alerts at the end. Those decide the running cost.
- Skofi's figures: Make, Zapier and CRM work US$25 an hour; custom systems from US$2,000; most US$2,000 to US$5,000; bigger builds US$5,000 to US$10,000. Quoted after discovery.
What does a custom integration cost?
It depends on the shape of the job, which is why any number given without a scope is a guess. We can state our own figures: Make, Zapier and CRM work is billed at US$25 an hour (US dollars), estimated in hours before we start. Custom systems start from US$2,000, and most projects come in at US$2,000 to US$5,000 depending on scope. Bigger builds with several features or an app run around US$5,000 to US$10,000. These are typical ranges, not a quote, and we quote after a discovery conversation. They describe Skofi, not the market, and we do not know what other firms charge.
For the cost of a specific kind of project, see our AI integration cost and automation cost articles. This one is about scoping the integration itself, so you can compare quotes and know what moved the number.
What drives the cost
| Driver | Why it matters | Cheaper side | Dearer side |
|---|---|---|---|
| Number of systems | Each system has its own API, objects, quirks and failure modes | Two systems | Three or more, or one through a hub |
| Direction | One-way needs one master. Two-way needs conflict rules for every shared field | One-way push or pull | Two-way with edits on both sides |
| Authentication | API key is simple. OAuth needs token refresh, consent flows and revocation handling | A single API key | OAuth per user, or several accounts |
| API quality | Documentation, sandbox availability, pagination, rate limits and change history | Well-documented, with a test environment | Thin documentation, no sandbox, or a product whose API differs by edition |
| Volume and timing | Records per day, bursts, and whether results are needed in seconds or by morning | Hundreds per day, overnight | Many thousands, or near real time |
| Data quality | Matching keys, duplicates, formats and missing values decide how much cleaning code is needed | Clean unique IDs on both sides | Free text matched by name |
| Error handling | Retries, idempotence so re-runs do not duplicate, and what to do with a record that will never succeed | Fail and tell a person | Automatic retry, dead-letter queue and recovery |
| Monitoring | Alerts, logs and a way to see what ran last night | A daily summary | A dashboard and escalation |
| Ownership and handover | Whose accounts, hosting and credentials, and who fixes it next year | Your accounts, notes included | A complex build only its author understands |
Where this breaks as a rule of thumb: a short field list can hide a hard integration. Ten fields synced two ways between a CRM and an accounting package with edits on both sides is harder than a hundred fields pushed one way every night.
No-code platform or custom code?
The first cost decision is whether the integration needs custom code at all. If both systems have good Make, Zapier or n8n modules and the volume is modest, a platform build is quicker and the next person can read it. Per-run billing becomes the concern as volume grows, so count the units on the vendor's current pricing page; our Make vs Zapier vs n8n article shows how.
Custom code earns its cost when volume is high, when logic goes beyond what modules allow, when you need exact control of retries and keys, or when the API is one no platform covers well. Vendor products with several editions are a typical case. MYOB, for example, has more than one product behind its API and features differ by product, so a custom integration there needs the exact product and version pinned down before it can be scoped. Our MYOB automation article covers that boundary.
A scoping checklist
Bring answers to these to any integration consultant. They turn a vague request into something that can be estimated, and a consultant who does not ask them is estimating blind.
- 01Which systems, which editions or plans, and do you have admin access to each?
- 02Which objects move (contacts, orders, invoices) and roughly how many per day or month?
- 03Which direction, and for two-way, which system is master for each field?
- 04How fast must a change appear in the other system: seconds, hourly, overnight?
- 05How are records matched, and what is the current state of duplicates and missing IDs?
- 06What counts as a failure, who should hear about it and what should happen to the record?
- 07Is there a sandbox or test account? If not, how do we test without touching live data?
- 08Who holds the credentials and the hosting, and who maintains it after handover?
- 09Are there compliance constraints on where the data may go? State them honestly; Skofi holds no SOC 2 or ISO certification and answers security questionnaires truthfully.
- 10What is the cost of it going wrong, and what is it costing you now?
Fixed price or hourly?
| Pricing | Fits when | Where it breaks |
|---|---|---|
| Fixed price | Scope can be written down: named systems, objects, direction, error rules, handover | When the API turns out different from its documentation. A good fixed quote states what is out of scope and what triggers a change |
| Hourly with an estimate | Discovery, an unfamiliar API, or work that grows as you learn what users need | No cap and no discipline. Ask for an estimate in hours before work starts and a check-in when it is exceeded |
| Paid discovery, then fixed | A large or uncertain build | You pay before you know the total, but the build quote is far firmer |
Whatever the structure, ask each bidder which variable would move their number. If they cannot name it, they have not understood the job. Our hiring a Make.com or Zapier expert article covers vetting in more detail.
The cost after launch
- Platform or hosting fees: Make, Zapier or n8n subscriptions, or a small server for custom code. Check the vendor's current pricing for your volume.
- Vendor API changes. Systems change their APIs, tokens expire, and a plan change can alter limits. Budget a little time for this each year.
- Someone to read the alerts. An integration with no owner is the cheapest thing to build and the most expensive to ignore.
- Change requests. New fields or a new system are the normal next step. A clean first build makes them cheaper.
We build this work under systems integration, including a fleet order build on monday.com described in our case study. Nothing in this article describes client results beyond what that page states.
Common questions
It depends on systems, direction, authentication, volume, error handling, monitoring and ownership. Skofi's own figures: Make, Zapier and CRM work at US$25 an hour, custom systems from US$2,000, most projects US$2,000 to US$5,000 depending on scope, bigger builds US$5,000 to US$10,000, quoted after a discovery conversation. These are ours, not a market average.
Scopes the integration, reads the vendors' API documentation, designs the data flow and the failure handling, builds or specifies it, and hands it over with notes and alerts. A good one also tells you when a no-code module or an existing connector is enough.
Usually, because every shared field needs a master or a conflict rule, and edits on both sides can overwrite each other. One-way flows have a single source and are simpler to test and recover.
Fixed price when the scope is writable, hourly with an hours estimate for discovery or unknown APIs. A paid discovery followed by a fixed build is a good option for larger work.
We build integrations for MYOB, as our MYOB automation article describes, and we are not a MYOB partner. Scope depends on the exact MYOB product and its API features, so that has to be confirmed first.
Platform or hosting fees, occasional fixes when a vendor changes its API, and someone to act on alerts. Ask any bidder to list these separately from the build price.
Drafted with AI assistance (Claude). Published under Saveez Irfan’s name, who is accountable for it. Editorial policy