26 October 2026
In short
How do you integrate an ERP without replacing it?
ERP integration in Coventry is usually about getting data in and out of a system nobody wants to replace. GTX Digital builds a layer beside the ERP rather than inside it, so the integration survives the vendor upgrade that would otherwise wipe out customisation.
- Build beside the ERP, never inside it. Customisation dies at the next upgrade.
- The ERP is the system of record. Everything else reads from it.
- Nightly batch is fine far more often than anyone admits.
- Ask the vendor what your support contract says about integration before you start.
- Most ERP pain is a reporting problem wearing an integration costume.
Think this applies to your business?
We will tell you straight whether it is worth doing, what it would take, and what we would do instead if it is not.
Stuck with an ERP that will not let go of your data? Book a technology review. We have worked beside most of the common ones and we will tell you what is realistically reachable.
Nobody in Coventry rings up wanting to replace their ERP. They ring up because the data is trapped inside it.
That is a different, much cheaper problem.
Beside it, not inside it
The first rule, and the one that saves the most money later.
Do not customise the ERP. Build alongside it.
Customisation inside the ERP feels direct, and it works beautifully until the vendor releases an upgrade. Then the customisation has to be reapplied, retested, and sometimes rewritten. Firms that have been through this twice end up never upgrading, which is how a business ends up running an unsupported version in 2027.
A layer beside the ERP reads through its API or supported views, does the clever part in your own code, and writes back the same way. The upgrade happens and your layer carries on.
Our ERP integration work is built on that principle, and the Coventry page for it sets out the engagement.
The ERP is the system of record
Decide this once and defend it.
Stock, orders, financials — the ERP is right. Every satellite system reads from it. Nothing else is allowed its own private version of a figure that lives there.
Where another system needs to add information, it adds information, it does not re-hold the master data. The warehouse app records a pick; it does not keep its own idea of stock.
Teams that skip this end up with four reports that each claim a different revenue number, and a finance meeting that is mostly archaeology.
Nightly is usually fine
Everyone asks for real time. Very little of a business needs it.
Real time means an event every time something changes, error handling for when the other end is down, and a queue so nothing is lost. It is more code, more monitoring and more to go wrong.
Nightly means one job, at a quiet hour, that you can re-run tomorrow if it fails.
So ask what breaks if the figure is twelve hours old. For management reporting, dashboards and most analysis, nothing breaks. For a customer portal showing whether an order has shipped, something does.
Mix the two. Nightly for the bulk, live for the handful that earn it.
Ask the vendor first
Before any work starts, find out what your support agreement actually permits.
The levels of access are not equally safe, and the difference is your support agreement.
| What you want to do | Usually permitted? | Why |
|---|---|---|
| Read the supported API | Almost always | It is what the vendor built it for |
| Read database views the vendor publishes | Usually | Published means supported |
| Read the underlying tables directly | Often tolerated, rarely supported | Nothing stops you, and nothing protects you either |
| Write to the underlying tables | Usually voids support, deservedly | The ERP writes things in its own order, for reasons |
Get the answer in writing from the vendor, not from the reseller who wants the work.
That one email has saved several of our clients from building something that would have been quietly unsupportable.
What people actually want
In practice, three things come up again and again.
| What they ask for | What it actually is | What it costs |
|---|---|---|
| Figures to look at without an ERP licence for every manager | A reporting job: nightly extract, somewhere queryable, a view on top | The smallest of the three, often weeks |
| To stop typing the same thing into two systems | A feed, and the only genuine integration of the three | The middle one, and the one worth scoping carefully |
| Customers to be able to see something | A portal in front of the ERP, not access to it | Depends entirely on what you expose |
Only the second of those is really an integration, and knowing which one you have changes the cost considerably.
The Coventry version
Engineering firms here tend to have an ERP chosen for manufacturing, and a set of requirements that arrived afterwards from customers.
A tier-one asks for an ASN in a particular format. Another wants a weekly file of open orders. A third has a portal of its own that wants uploading to.
None of those are things the ERP shipped with. They are integrations, each small, and the trap is building three unrelated ones. Build the extraction once, then three small adapters for the three formats. Adding a fourth customer is then a day, not a project.
That pattern matters here because the customer list is long and each one has an opinion. It is the same shape of work we do across manufacturing and the automotive supply chain.
The data is dirtier than anyone admits
Every ERP integration meets this on about day three.
It is always the same list, in a different order.
- Part numbers with trailing spaces.
- The same supplier under two codes, because somebody set one up in a hurry.
- Units of measure that are "each" in one place and "EA" in another.
- Dates stored as text.
- A field everyone agreed to stop using in 2021 that half the records still rely on.
None of that is a scandal. It is what twenty years of a working business looks like from the inside.
But it has to be found early, because it changes the estimate. Write a throwaway script in the first week that profiles the tables you intend to use: how many nulls, how many distinct values, how many rows break the rule everyone assured you was always true.
That profile is worth more than a fortnight of requirements workshops. It tells you which assumptions are safe and which need a cleaning job before anything else happens.
Budget for the cleaning as its own piece of work with its own number. Hiding it inside the integration estimate is how projects end up looking late when they were only ever optimistic.
When replacement is the honest answer
Sometimes the ERP is genuinely the problem. Unsupported version, vendor gone, no API, nobody left who understands it.
Even then, do not replace it in one go. The route is the same as any other: slice it, run old and new side by side, move one area at a time. That is the subject of legacy software replacement in Coventry.
Replacing an ERP because reporting is hard is an expensive way to buy a report.
Where to start
List the five numbers your managers ask for most often, and where each one lives today.
If most of them live in the ERP and the pain is getting them out, you have a reporting problem and it is solvable in weeks.
If they live in four places and disagree, read CRM integration in Coventry first, because ownership has to be settled before anything is wired together.
Our work across the city is on the Coventry page. When you want it scoped properly, talk to us.
Common questions
Can we integrate without upgrading the ERP?
Usually yes. A layer alongside it, reading and writing through supported routes, avoids touching the ERP itself.
Does integration void ERP support?
Modifying the ERP often does. Integrating through its supported API or database views normally does not. Get it confirmed in writing.
Real time or overnight?
Overnight is cheaper, simpler and sufficient for most reporting. Keep real time for the few things where minutes genuinely matter.
What about the next upgrade?
An integration that only uses supported interfaces survives it. One that relies on undocumented tables does not.
Next step
Start with a technology review, not a proposal.
A short, structured conversation about where your systems are costing you time, margin or growth — and what the realistic options are. No pitch deck.