ERP integration cost is the line in the business case that moves the most and gets questioned the least. The licence is a published number, the implementation partner quotes a day rate, and then the connections between the new system and everything else in the business are summarised in one word — “interfaces” — with a figure underneath it that nobody can defend. That figure is where projects go wrong, because the work it stands for is the least understood part of the programme.
This guide takes that line apart. It covers what ERP integration cost actually includes, which factors move it by a factor of five, how the main integration approaches compare on total spend rather than day one, what realistic price bands look like for UK companies of different sizes, where the first-year budget really goes, the costs most quotes quietly omit, and a phased plan that keeps the number under control without cutting the scope that matters.
Every figure here is a planning model rather than a quotation. The assumptions sit next to the numbers so you can replace them with your own and see what changes. If you are earlier in the decision than this, our build vs buy vs low-code decision framework covers the choice that comes before integration planning, and the companion guide to API integration cost goes deeper on individual connections.
Table of contents
- What ERP Integration Cost Actually Covers
- What Drives ERP Integration Cost Up or Down
- ERP Integration Cost by Approach: Point-to-Point, Middleware and iPaaS
- Realistic ERP Integration Cost Bands for UK Businesses
- Where the First-Year ERP Integration Cost Actually Goes
- The Hidden Costs Most Quotes Leave Out
- Data Migration and Master Data: The Line Item That Sinks Budgets
- A Phased Plan That Controls ERP Integration Cost
- Running Costs After Go-Live
- How to Reduce ERP Integration Cost Without Cutting Scope
- Frequently Asked Questions
- References
What ERP Integration Cost Actually Covers
Before anyone can budget for it, everyone in the room needs the same definition of what is being bought.
The scope most quotes silently assume
An ERP integration cost line usually covers the design, build, testing and deployment of the data flows between the ERP and the systems that surround it. It rarely covers cleaning the data those flows will carry, redesigning the processes they will automate, or supporting them after go-live. Three of the five real workstreams sit outside the number, which is precisely why the number looks affordable.
Integration is not part of the ERP licence
Vendors sell platforms; connections are a project. A subscription to a modern ERP entitles you to APIs, and in most cases to a set of published connectors — it does not entitle you to a working link between your ERP and your warehouse management system. Treating the licence as if it includes integration is the single most common budgeting error we see in early business cases.
The five categories every estimate should name
A defensible ERP integration cost estimate breaks into discovery and mapping, build and configuration, data migration, testing and parallel running, and cutover with hypercare. If a proposal shows one number and no breakdown, ask for these five. A partner who cannot split them has not scoped the work, and the eventual invoice will reflect that rather than the quote.
Why the same project gets very different prices
Two proposals for identical scope can differ threefold because they assume different things about your data, your availability, and how much process change you will accept. One assumes your product master is clean; another has priced two weeks of deduplication. Neither is dishonest. Comparing them without normalising the assumptions is what makes the ERP integration cost debate so unproductive.
What Drives ERP Integration Cost Up or Down
A small number of variables account for most of the variance, and only one of them is technical.
Connections, not headcount
ERP integration cost scales with the number of connected systems and the number of distinct data objects crossing between them, not with how many people log in. A 40-person distributor with an e-commerce store, a warehouse system, a courier platform and a payment gateway will pay more than a 400-person firm running finance and payroll alone. Count interfaces before you count seats.
The state of your existing data
Data quality is the dominant multiplier. If customer records exist in three systems with three different keys, someone has to decide which is authoritative and reconcile the rest before any interface can be trusted. Our AI-ready data checklist covers the same discipline for a different purpose, and the underlying work is identical.
Whether a supported connector already exists
A published, vendor-maintained connector between two mainstream platforms can turn six weeks of build into four days of configuration. A bespoke connection to a twenty-year-old system with a file drop and no documentation goes the other way. Before estimating anything, list each connection and mark it supported, semi-supported or bespoke. That single column explains most of the spread in any ERP integration cost model.
Real-time versus batch
Real-time synchronisation costs materially more than an overnight batch, because it demands error handling, retry logic, idempotency and monitoring that a nightly file simply does not. A lot of real-time requirements dissolve when someone asks how quickly the business genuinely needs the record to appear. Frequency is a commercial decision dressed as a technical one.
Process change hiding inside the interface
The most expensive integrations are the ones where two systems disagree about how the business works. Reconciling a sales order model with a manufacturing work order model is not a mapping exercise, it is an operating decision, and it needs the people who own the process rather than the people who own the middleware.
ERP Integration Cost by Approach: Point-to-Point, Middleware and iPaaS
The architecture you choose changes both the initial bill and the shape of the spend over five years.
Point-to-point: cheapest to start, dearest to keep
Direct connections between pairs of systems are quick and cheap for the first two or three. The problem is combinatorial: every new system multiplies the links, and every upgrade risks several of them at once. Point-to-point keeps year-one ERP integration cost low and quietly builds a maintenance liability that arrives in year three.
Native and vendor-supplied connectors
Where the ERP vendor publishes a connector to a system you already run, use it. You inherit the vendor’s testing, upgrade compatibility and support, and you pay in configuration time rather than development time. The constraint is that connectors implement the vendor’s view of the process, so anything unusual still needs custom work around the edges.
Middleware and enterprise service buses
A middleware layer centralises routing, transformation, logging and error handling. It costs more up front and pays back once you pass roughly five connected systems, because each additional system connects once to the bus rather than separately to everything else. It also gives you one place to look when a record goes missing, which is worth more than most business cases credit.
Integration platform as a service
An iPaaS converts a large part of the build into a subscription. Prebuilt connectors, a visual mapping surface and managed infrastructure reduce the initial spend and the specialist skills needed, in exchange for a recurring fee that never stops and a platform you are now dependent on. For mid-market firms without an internal integration team, this is frequently the lowest total ERP integration cost over five years.
| Factor | Point-to-point | Native connector | Middleware/ESB | iPaaS |
|---|---|---|---|---|
| Initial build cost | Low | Lowest | High | Moderate |
| Recurring platform fee | None | In licence | Moderate | High |
| Cost of system six | Very high | n/a | Low | Low |
| Upgrade fragility | High | Low | Moderate | Low |
| Error visibility | Poor | Moderate | Excellent | Good |
| Skills required | Developers | Functional | Specialists | Mixed |
| Time to first flow | 2-4 weeks | 3-10 days | 6-12 weeks | 2-5 weeks |
| Best at | 2-3 systems | Mainstream pairs | 8+ systems | 4-10 systems |
How the approaches diverge over five years
Year one flatters point-to-point and punishes middleware. By year five the ordering usually reverses, because maintenance, upgrade regression and each new connection all land on the architecture rather than on the original build.
Realistic ERP Integration Cost Bands for UK Businesses
Numbers are more useful than principles, provided the assumptions travel with them.
The assumptions behind these bands
Each band assumes a UK delivery partner at mainstream mid-market rates, a cloud ERP with documented APIs, hosting on mainstream cloud computing platforms at list prices, data that is imperfect but not chaotic, and a business that can release subject-matter experts for testing. Change any of those and the band shifts. Public sector procurement, multi-currency consolidation and regulated industries all sit above these figures.
Small business: one ERP, two or three connections
A company connecting finance to a payments provider and an e-commerce store, with modest volumes and standard processes, is typically looking at £18,000 to £45,000 of ERP integration cost. Most of it is discovery, mapping and testing rather than code, which surprises people who expected the build to dominate.
Mid-market: finance, CRM, warehouse and commerce
Four to seven connections with real transaction volume and some bespoke logic lands between £70,000 and £180,000. This is the band where architecture choice starts to matter more than day rate, and where an iPaaS or a middleware layer usually justifies itself against a growing tangle of direct links.
Enterprise and multi-entity
Multiple legal entities, consolidated reporting, warehouse automation, EDI with trading partners and a legacy system that cannot be retired yet will run from £250,000 upwards, often well beyond. At this size the ERP integration cost is dominated by coordination and testing, not development, and the programme needs an integration architect rather than a lead developer.
| Scenario | Connections | Indicative range | Elapsed time | Biggest risk |
|---|---|---|---|---|
| Small business | 2-3 | £18k-£45k | 6-12 weeks | Undocumented legacy data |
| Growing SME | 3-5 | £45k-£90k | 3-5 months | Scope creep in mapping |
| Mid-market | 4-7 | £70k-£180k | 5-9 months | Master data conflicts |
| Multi-site manufacturer | 6-12 | £150k-£400k | 9-15 months | Shop-floor downtime |
| Multi-entity enterprise | 10+ | £250k+ | 12-24 months | Coordination and testing |
| Regulated or public sector | Varies | Add 25-40% | Add 3-6 months | Assurance and audit |
What a suspiciously cheap quote usually means
A quote well below these bands is normally excluding data migration, assuming your team writes the mapping specifications, or pricing a single happy-path flow with no exception handling. None of that is fraud; it is an unstated assumption. Ask what happens when a record fails validation at three in the morning, and the real ERP integration cost appears in the answer.
Where the First-Year ERP Integration Cost Actually Goes
The intuition that integration is mostly coding is wrong, and budgeting on that intuition is why so many projects run out of money in testing.
Discovery, mapping and specification
Field-level mapping between systems, agreeing the system of record for every object, and documenting the exception rules typically absorbs a fifth of the budget. It is unglamorous and it is the work that determines whether the build goes smoothly. Skipping it does not save money; it moves the money into rework.
Build and configuration
Actual construction is usually a quarter to a third of first-year ERP integration cost. Where supported connectors exist this shrinks dramatically, which is why the connector inventory matters more than the day rate when you are comparing proposals.
Data migration and cleansing
Extracting, deduplicating, transforming and loading historical data is routinely the largest single item and the most frequently underestimated. Our guide to legacy system modernisation covers the wider decision about what to carry forward at all.
Testing, UAT and parallel running
Integration testing needs both systems, realistic volumes and people who know what a correct answer looks like. Parallel running — processing the same period in both the old and new systems and reconciling the difference — is expensive and is the only test that reliably finds the problems that matter.
Cutover and hypercare
The switchover weekend plus four to six weeks of elevated support is a real budget line, not a rounding error. Plan for a temporary drop in throughput while people learn the new flows, and staff the first month accordingly.
These items are rarely dishonest omissions. They are simply outside the scope a systems integrator is asked to price.
Licence tiers and API consumption
Many platforms meter API calls or gate integration features behind a higher tier. A design that polls every five minutes can quietly move you up a pricing band, and the finance team discovers it a quarter later. Check the call limits before you agree the synchronisation frequency, not after.
Environments, sandboxes and refreshes
You need at least a development and a test environment that resemble production, and you need to refresh them with realistic data more than once. Each environment carries a licence and each refresh carries effort. On a mid-market programme this is commonly five to eight per cent of the ERP integration cost.
Your own people’s time
The internal cost is usually the largest unbudgeted item. Subject-matter experts writing rules, testing scenarios and reconciling parallel runs are doing a second job for months. Costing their days at a notional rate makes the trade-offs visible and stops the programme assuming infinite availability.
Customisations that break at upgrade
Every bespoke field and every custom transformation becomes a regression test on the vendor’s next release. Two release waves a year, multiplied across a dozen interfaces, is a standing annual commitment — the same dynamic our software maintenance cost guide describes for bespoke applications.
Exception handling and reconciliation tooling
Somebody has to see failed records and fix them. A queue, an alert, a retry mechanism and a reconciliation report are a small project in their own right, and leaving them out is the most reliable way to make a cheap integration expensive in its first month of live operation.
| Hidden item | Typically excluded because | Budget impact |
|---|---|---|
| API tier upgrade | Vendor commercial, not project scope | 3-8% annually |
| Extra environments | Assumed already available | 5-8% one-off |
| Internal SME time | Not a supplier invoice | 15-30% equivalent |
| Data cleansing before load | Client responsibility in the contract | 10-25% one-off |
| Exception and retry tooling | Outside the happy path | 6-12% one-off |
| Upgrade regression testing | Falls after go-live | 8-15% annually |
| Trading partner onboarding | Depends on third parties | £2k-£6k each |
Data Migration and Master Data: The Line Item That Sinks Budgets
If one workstream is going to overrun, this is the one, and it overruns for reasons that have nothing to do with technology.
Master data is a governance problem
Deciding that the CRM owns the customer record and the ERP owns the invoice is an organisational decision with political weight. Until it is made and written down, every interface design meeting reopens it. The technical work is trivial by comparison; the agreement is what takes weeks.
Deduplication consumes the days
Merging three customer lists is not a script, it is a series of judgement calls about which duplicates are genuinely the same entity. Expect a human in the loop, expect disagreement, and expect the count of “records requiring manual review” to be larger than anyone predicted at kick-off.
History: how much do you genuinely need?
Migrating ten years of transactions because nobody wanted to make a decision is a common and expensive default. Two years live plus an archive that finance can query satisfies most requirements at a fraction of the effort, and it reduces ERP integration cost immediately. The ICO’s guidance on data minimisation is a useful lens: keeping less is both cheaper and more defensible.
Reconciliation is the real acceptance test
The migration is finished when the trial balance matches, the stock valuation matches and the open order book matches — not when the load script completes without errors. Write those reconciliation criteria into the contract, because they are the only objective definition of done.
A Phased Plan That Controls ERP Integration Cost
Sequencing does more for the budget than negotiation does. This is the plan we recommend to clients regardless of platform.
Phase 0: build the integration inventory
List every system, every data object that must cross a boundary, the direction, the frequency and the system of record. One page per interface. This artefact alone typically changes the ERP integration cost estimate by a third, usually downwards, because it exposes flows nobody actually needs.
Phase 1: one connection, end to end
Deliver a single interface completely — built, tested, monitored, documented and running in production. It calibrates every subsequent estimate against your real conditions rather than a partner’s standard assumptions, and it surfaces environment and access problems while they are still cheap.
Phase 2: finance first
Finance interfaces have the clearest rules, the most objective tests and the loudest stakeholders. Getting the general ledger, accounts payable and receivable flows correct early gives the programme credibility and makes reconciliation possible for everything that follows.
Phase 3: the operational spine
Orders, stock and fulfilment come next, once the financial picture is trustworthy. This is the phase where process change is real, so pair every interface with the operational owner who has authority to change how the work is done.
Phase 4: the long tail
Reporting extracts, partner feeds and the handful of departmental tools finish the programme. Deliberately schedule some of them after go-live: several will turn out to be unnecessary once the core is live, and deferring them is a legitimate way to reduce ERP integration cost without losing anything the business values.
Running Costs After Go-Live
Integration is not a capital project that ends. Budgeting it as one guarantees an unpleasant conversation in year two.
Platform subscriptions and consumption
An iPaaS or middleware licence recurs, and consumption-based pricing rises with your transaction volume. Model it against your growth plan rather than today’s volumes, because success makes this line grow faster than any other.
Vendor release cycles
Mainstream cloud ERPs ship major updates twice a year and do not let tenants defer indefinitely. Each wave means regression testing your interfaces. Treat it as a scheduled annual cost with named owners, in the same way you treat an audit.
Monitoring, alerting and someone on call
A failed overnight sync that nobody notices until Tuesday is more expensive than the monitoring that would have caught it. Alerting, a dashboard and a documented escalation path are modest costs that prevent expensive ones.
The annual percentage to budget
As a planning figure, budget annual running costs at 15 to 25 per cent of the original build for a middleware or iPaaS estate, and rather more for a point-to-point tangle, where every change ripples. This is the single most useful number to put in a five-year model.
How to Reduce ERP Integration Cost Without Cutting Scope
Every tactic here removes work rather than quality, which is the only kind of saving that survives contact with go-live.
Standardise the process before you automate it
Automating an inconsistent process means building the inconsistency into code and paying to maintain it forever. Three regional variants of the same approval flow become three interfaces. Agreeing one variant first is the largest single reduction in ERP integration cost available to most organisations, and it costs meeting time rather than money.
Refuse the real-time you do not need
Challenge every synchronous requirement. If the warehouse picks twice a day, a two-hourly batch is indistinguishable from real time in outcome and materially cheaper in build, monitoring and support. Reserve genuine real-time for the flows where a delay has a commercial consequence.
Buy the connector, build only the exception
Use supported connectors for the standard 80 per cent and reserve bespoke development for the genuinely distinctive remainder. This keeps the upgrade surface small, which is where multi-year ERP integration cost actually accumulates.
Contract for outcomes, not for days
Define acceptance as reconciled data and a working flow in production, not as a number of consultant days delivered. Our guidance on choosing between Microsoft Dynamics 365 and SAP covers how platform choice interacts with the commercial model, and the principle holds either way.
Own the mapping documents yourself
Insist the field mappings, transformation rules and exception logic are delivered as your documentation, in your repository. It costs nothing during the project and it is what makes the second phase, the next partner or the eventual replatform affordable rather than a rediscovery exercise.
Frequently Asked Questions
How much should we budget for ERP integration cost as a share of the ERP project?
For a mainstream cloud ERP, integration commonly accounts for 25 to 45 per cent of total programme spend, and the share rises with the number of surrounding systems. Below 20 per cent, check whether data migration and testing have actually been priced.
Is an iPaaS cheaper than building integrations ourselves?
Usually yes over five years for four or more connections, because you avoid specialist hiring and inherit maintained connectors. Below three connections, the subscription rarely pays for itself and direct connections are the pragmatic answer.
How long does ERP integration take?
Six to twelve weeks for two or three straightforward connections, five to nine months for a typical mid-market estate, and a year or more for multi-entity programmes. Data quality moves these ranges more than technology choice does, and our software development timeline guide explains why estimates slip.
Can we phase integration after the ERP goes live?
Yes, and it is often sensible. Launch with finance and the one or two operational flows the business cannot run without, then add the rest. Plan the interim manual process explicitly rather than discovering it in week one.
What is the most underestimated part of the budget?
Data migration and the internal time it consumes. It is the largest single item in most first-year models, the hardest to scope in advance, and the one that sits in a contract clause labelled “client responsibility”.
Does AI reduce ERP integration cost?
Modestly and unevenly. Assisted mapping and test generation genuinely shorten build tasks, but the expensive parts — deciding the system of record, reconciling process differences, agreeing what a correct answer looks like — remain human decisions that no model can make for you.
References
Microsoft Learn: Data Entities Overview for Finance and Operations
Microsoft Learn: Dual-Write Overview
Microsoft Learn: What Is Microsoft Dataverse
Microsoft Learn: Azure Logic Apps Overview
SAP Help Portal: SAP Integration Suite Documentation
ICO: Data Minimisation Under the UK GDPR
