What System Integration Actually Costs in Australia
The honest answer is that it depends, and anybody who gives you a number before asking what you are connecting is guessing. But the things it depends on are knowable, and the shape of the pricing matters far more than the headline figure. A cheaper build on the wrong pricing model can cost several times a dearer one within two years.
This page sets out the five pricing models you will meet, how each behaves when your business grows, the seven things that genuinely move the cost of a build, the ongoing spend almost nobody budgets for, and a way to compare two quotes that describe the same work in completely different words. All figures are in Australian dollars and excluding GST unless stated.
Realistic ROI
Four Things That Decide What You Will Pay
Before any specific number is useful, these four decide which order of magnitude you are in.
How many flows, not how many systems
People instinctively count systems, but cost tracks the number of distinct data flows and the direction of each. Connecting two systems with one flow in one direction is a small job. Connecting the same two systems with orders going one way, stock coming back, customers matched in both directions and refunds handled as a special case is four or five pieces of work with their own rules, exceptions and tests. When you ask for a quote, list the flows rather than the logos, and you will get comparable numbers back.
How messy the existing data is
This is the single most underestimated driver, and it is invisible until someone looks. Duplicate customers, products with no consistent identifier, inconsistent units of measure, a decade of historical records that predate the current rules, and free text fields where a code should be all turn a clean mapping exercise into a data remediation project. A supplier who has looked at your actual data before quoting will give you a number you can rely on. One who has not will give you a number that changes.
How much the failure of a flow costs you
An integration that moves marketing contacts and one that moves financial transactions cost different amounts to build correctly, because the second one needs reconciliation, audit logging, idempotent writes and a recovery path that the first can reasonably do without. It is legitimate to buy a cheaper standard of engineering for a low consequence flow. It is expensive to buy it for a high consequence one and find out at the end of a quarter.
Whether anyone is going to look after it
A build with no maintenance arrangement is not cheaper, it is deferred. Vendor interfaces change, authentication methods are retired, volumes grow past what was tested, and the person who wrote it moves on. The cost either appears as a monthly fee you agreed to, or as an emergency at a bad moment plus the cost of somebody unfamiliar reverse engineering an undocumented system. Comparing a quote that includes support against one that does not is not a price comparison at all.
The Five Pricing Models and How They Behave
The same project can be priced any of these ways. What matters is what happens to your bill in the third year and in your best trading month.
Fixed build plus a flat managed fee
A defined price for the build against an agreed specification, then a flat monthly fee that covers hosting, monitoring, vendor interface changes and a defined amount of support. Predictable in both directions, which is why it suits a business that wants to budget. The thing to check is precisely what the monthly fee includes, particularly whether responding to a vendor changing their interface is covered or billed as a new project, and what happens when you want a new flow added.
Time and materials
You pay for hours at a day rate. Honest and appropriate for genuinely exploratory work such as integrating an obscure system nobody has touched, and uncomfortable for a well understood build because all the estimating risk sits with you. If you are quoted this way for a standard flow, ask why. A reasonable middle ground is time and materials for a short discovery phase, producing a specification, followed by a fixed price for the build itself.
Per record, order or transaction
Common with platform vendors and superficially attractive because it starts small. The problem is structural: your bill rises exactly when trading is strongest, so a successful November costs you twice, and a returns heavy period can push volume up while margin falls. Model it against your actual peak volumes rather than your average, look for the point where it exceeds a flat fee, and check whether failed or retried records are counted as billable.
Platform licence plus implementation
An integration platform subscription, priced by connectors, tasks or volume, plus the cost of somebody configuring it. Genuinely good value when you have many simple connections and in house capability to run them. Watch for connector counts that force an upgrade tier, tasks metered in a way that penalises the polling your flows need, and the assumption that the platform fee replaces the implementation cost when it usually sits alongside it.
Bundled into another project
Integration folded into an ERP or platform implementation as a line item with no detail. The risk is that it is scoped generously enough to win the deal and specified thinly enough to be argued about later, and that when it runs over you have no independent basis to push back. Ask for the flows to be itemised even inside a larger contract, and ask who supports them after the implementation partner has finished and moved on.
The cost of doing nothing
The comparison that is almost always missing. Count the hours spent re-keying at a fully loaded hourly cost, then add the errors: oversells, wrong prices published, invoices raised twice, refunds missed, and the reporting decisions made on numbers that were three days stale. Most businesses find the manual figure is larger than they assumed and, importantly, that it grows with volume while an integration mostly does not. That comparison is what makes a business case defensible rather than a matter of taste.
What Different Kinds of Work Typically Involve
| Task | Traditional | What to Expect | Notes |
|---|---|---|---|
| One flow, one direction, common systems | Weeks of manual entry | Smallest project size | Orders into accounting between two mainstream cloud products. Often a supported connector does it for less than any build. |
| Two way sync between two systems | Double entry both ways | Several times the single flow | Ownership rules, conflict handling and loop prevention are most of the extra work, not the connection. |
| Hub connecting five or more systems | Point to point sprawl | A programme, not a project | Priced per flow and staged. The hub itself is cheaper than the eleventh point to point link. |
| A system with no usable interface | Manual forever | Materially more expensive | File drops or screen level automation are slower to build and more fragile to run. Sometimes the honest answer is to change the system. |
| Historical data backfill | Quoted as an afterthought | Priced separately | Years of imperfect records is a data project with its own testing. Ask for it as a line item. |
| Ongoing operation and monitoring | Assumed to be free | A recurring fee | Hosting, alerting, vendor changes and support. This is the line that decides three year cost. |
| Adding a flow later | A new project each time | Cheaper on a hub | Worth asking the price of the second and third flow before you sign the first. |
| Getting your integration back | Not discussed | An exit clause | Documentation, credentials and mappings handed over. Ask before signing, not when leaving. |
Where Integration Budgets Go Wrong
The quote priced a connection, not a solution
A number that covers moving data from A to B but not field mapping, matching rules, exception handling, testing on real data, documentation or monitoring is not a cheaper version of the same thing. It is a smaller piece of work that leaves the rest for you. When comparing quotes, write down the ten things you expect to receive and mark each supplier against them. Differences that look like price are usually differences in scope, and finding that out during the comparison is much cheaper than finding out during delivery.
Per record pricing was modelled on average volume
Averages hide the problem. Model any volume based pricing against your busiest month, then against that month plus the growth you are planning for, then against a returns heavy period where record counts rise while revenue does not. Ask explicitly whether retries, failures, polling checks and internal steps count as billable records, because the difference between counting orders and counting operations can be an order of magnitude. Then work out at what volume a flat fee becomes cheaper, and whether you expect to cross it.
Data remediation was nobody’s line item
Almost every integration meets duplicate customers, inconsistent product identifiers and historical records that do not follow current rules. If the quote does not mention cleaning any of it, either the supplier has not looked at your data or they intend to raise it as a variation later. Ask for a data assessment before the build is priced, ask what happens to records that cannot be matched, and expect the answer to include a review queue for a person rather than a promise that everything will map.
The business case ignored the run cost
Building an integration and then leaving it unmonitored and unmaintained produces an asset that degrades. Interfaces change, certificates expire, volumes grow past what was tested, and the failure is usually silent. Whatever the arrangement, the three year figure has to include hosting, monitoring, someone responding when a vendor changes something, and a realistic allowance for small changes. A build only quote is not the cheaper option, it is the incomplete one.
Nothing was written down about ownership
Ask before you sign who owns the integration logic, the field mappings and the documentation, and what you receive if the relationship ends. Some arrangements leave you with nothing but a working system you cannot see inside, which turns switching suppliers into a rebuild and gives you no leverage at renewal. A reasonable arrangement gives you the specification, the mappings, the system map and your own credentials, whether or not you are the one operating it day to day.
The cheapest quote was the least specified
There is a reliable pattern where the lowest number comes with the shortest scope, and the gap is closed later through variations at a rate you did not negotiate. The defence is to write the specification yourself, or have somebody independent write it, and ask every supplier to price the same document. It costs a little up front and it converts a set of incomparable proposals into an actual comparison, which is usually worth more than the difference between the quotes.
How Yes AI Prices Integration Work
Scoping before pricing, always
We map the systems, look at real data samples and count the manual hours before quoting. If that work shows the project is smaller than you feared, or that a supported connector covers it, you get that answer rather than a proposal. The map and the recommendations are yours regardless of what you do next.
Fixed build price and a flat managed fee
A fixed price against an agreed written specification, then a flat monthly fee covering hosting on a managed cloud automation layer we operate, monitoring, and keeping the integration working when a vendor changes their interface. No per order or per record charging, so your best trading month is not also your largest bill.
The documentation is a deliverable
System map, ownership decisions, field mappings and runbooks are handed over and kept current as the integration changes. If you ever want to run it yourself or move it elsewhere, you have what you need. We would rather earn the renewal than rely on you being unable to leave.
A three year figure, not just a build price
Every proposal shows build, run and the cost of the manual work it replaces over the same period, so the comparison is honest. Where we think the payback is weak or the timing is wrong, we say so before you commit rather than after.
How to Get a Number You Can Trust
Five steps that turn a vague budget question into a decision you can defend to a board.
Count the manual work in dollars
Time the tasks people repeat, apply a fully loaded hourly cost, and add the measurable errors: oversells, duplicate invoices, missed refunds, stale reporting. This is the baseline every quote should be compared against, and it is usually larger than expected.
List the flows, not the systems
Each data type, its direction, its frequency and which system owns it. A one page list of flows is the single most valuable thing you can bring to a pricing conversation, because it makes different suppliers quote the same thing.
Look at the data before anyone quotes
Sample the customer, product and transaction records and count the duplicates, the missing identifiers and the historical oddities. If a supplier will not look at this before pricing, treat their number as provisional whatever the covering letter says.
Compare on three years, on the same scope
Build plus run plus expected changes, on identical scope, with volume based models stress tested against your peak month. Mark every quote against the same checklist of deliverables so scope differences do not disguise themselves as price differences.
Agree the exit before you agree the start
Who owns the logic, mappings and documentation, what you receive if you leave, how quickly credentials can be rotated, and what the second and third flows will cost. Settle this while you have negotiating power rather than when you need it.
Related Reading
SaaS Integration Explained
The six patterns, what each costs and how to choose.
iPaaS in Australia
Where a platform licence earns its keep and where it does not.
Writing the Specification
The document that makes quotes comparable.
Questions to Ask a Vendor
Checking integration capability before you buy the software.
Monitoring and Support
What the ongoing fee should actually cover.
Custom API Integration
When the connector you need does not exist yet.
FAQ
Get a Number You Can Take to the Board
Book a call. We map the systems, sample the data and count the manual hours, then give you a scoped, priced plan with a three year view. If a supported connector does the job for less, we will tell you that instead.
All discussions held in confidence. Australian-based consultants.