- Market
- Saudi Arabia
- Written for
- CIO, COO
- Decision
- Understand why department systems disagree and how to connect them without a platform overhaul
The shape of the problem
Five systems, five customers, one company
Each system below is right by its own records and wrong by everyone else’s. The middle column is not a product to buy; it is a set of decisions about ownership, made once and enforced by whatever integration is then built.
A surface diagram: five departmental systems — CRM, ERP, storefront, warehouse, support — each holding its own copy of core records, converging on an ownership-and-integration layer that produces one customer record, one order status and reports that agree.
Sales CRM
Its customer: the contact the salesperson knows, with the deal history.
Finance ERP
Its customer: the legal entity that gets invoiced, with the balance.
E-commerce storefront
Its customer: the account that logs in, with the cart and addresses.
Warehouse system
Its order: what was picked and shipped, on its own numbering.
Support desk
Its customer: whoever wrote in, matched by email if at all.
What we do
Ownership decided, then enforced
For each record type, one system is named the owner; the others hold copies they receive, not versions they invent. Integration then carries changes from the owner outward.
One customer record
Sales, finance and support describe the same party.
One order status
What was sold, shipped and billed reconciles without a spreadsheet.
Reports that agree
Two departments answering the same question get the same number.
The integration is plumbing. The ownership decision is the fix.
The daily cost
The re-keying tax is paid in salary and settled in errors
When systems do not talk, people become the integration. An order confirmed in the storefront is typed into the ERP; a customer updated in the CRM is updated again, or not, in the billing system; month end becomes a reconciliation exercise between exports. This tax has two parts. The visible part is hours — skilled staff doing transcription. The expensive part is the error rate that comes with every manual copy: a digit transposed in a price, a delivery going to last year’s address, a report built on the copy that missed an update. Nobody chose this workload; it condensed, order by order, out of the gaps between systems.
The deeper fight, though, is not about typing — it is about authority. Ask which system owns the customer and watch the meeting stall: sales says the CRM, finance says the ERP, and both are right about their own slice. Master data is a governance question wearing a technical costume, and it does not have a neutral answer; someone senior has to rule, per record type, whose version wins. Organisations that skip this ruling and go straight to integration tooling discover that they have automated the argument: the systems now overwrite each other’s versions at machine speed.
The honest options
Point-to-point, middleware, or a staged programme — what each really costs
None of these is wrong in itself. Point-to-point gets a bad name from being used past its limits; middleware gets oversold as a cure for governance problems it cannot touch. The number of systems and the rate of change decide, not fashion.
| Option | How it works | Where it is honest | Where it breaks down | Best for |
|---|---|---|---|---|
| Point-to-point | Each pair of systems that must share data gets its own direct connection, built on whatever interfaces the two expose. | Two or three systems, stable, with one or two well-understood flows between them. Cheapest and simplest by far at that scale. | Every added system multiplies connections; a change in one system ripples into every link it holds, and eventually nobody can say what talks to what. | A small, stable estate with a couple of clear flows. |
| Middleware / integration layer | Systems connect once to a shared layer that routes, transforms and monitors the flows between them. | Many systems, frequent change, and a need to see and audit the flows in one place. The layer earns its cost by being the single place a change is made. | When it is bought as a substitute for the ownership decision. A hub that faithfully synchronises five contradictory versions of the customer delivers contradiction as a service. | A growing estate where flows change often and need one point of control. |
| Staged programme | Ownership is ruled first, then interfaces are built in pain order — the worst manual bridge automated first, the architecture allowed to grow from point-to-point toward a layer as the count justifies it. | Most real estates: it delivers relief within the first stage instead of after a platform programme, and every later stage builds on ruled ownership. | It needs someone to hold the sequence. Without an owner, stage one ships and the programme quietly stops — leaving the estate better but the map unfinished. | An estate in active pain that cannot wait for a big-bang platform. |
Our role
Why the first interface we build is the one that hurts most
An integration programme earns trust by removing a pain someone feels weekly, not by presenting an architecture diagram. So we sequence by pain, and we make the governance decisions explicit instead of burying them in mappings.
- 01
The estate, inventoried
We list the systems, the copies of each core record, and every manual bridge between them — with the hours it consumes and the errors it produces — so the worst interface is identified by evidence rather than by whoever complains loudest.
- 02
Ownership put to the right table
We frame the master-data decisions — which system owns the customer, the product, the order — as the business rulings they are, prepare the options and consequences, and get them decided before a line of integration is built on top.
- 03
Relief shipped in stages
The most painful bridge is automated first and proven in daily use; each following stage extends the same ownership rules to the next interface. The architecture grows toward a layer when the count of systems justifies one, not before.
Whose version of each record wins is a ruling only your leadership can make; we prepare it honestly, then build the plumbing that enforces it — starting where your people lose the most hours today.
Common questions
What buyers ask
Can we just connect our two worst systems directly, or do we need middleware?+
If the pain is concentrated in one flow between two systems, a direct connection is the honest answer — it is faster to build, simpler to run, and does not commit you to a platform. Middleware earns its cost when the count of systems and the rate of change make direct links unmanageable: many flows, frequent changes, and a need to see them in one place. The pragmatic path is to fix the worst interface directly now, decide record ownership as you do it, and let the architecture grow into a layer if and when the estate’s size demands one. Buying the platform first inverts that: months of foundation work before anyone feels relief.
Which of our systems should own the customer record?+
The one where the record is born or legally anchored for your business — and that differs by company, which is exactly why no vendor can answer it for you. A useful test: for each field, ask which team is accountable when it is wrong. Billing details point at finance and the ERP; contact and relationship history point at sales and the CRM. Many organisations settle on split ownership — the ERP owns the legal and billing identity, the CRM owns the relationship — with the integration carrying each owner’s fields outward as read-only copies. What matters less than the particular ruling is that there is one, written down, per record type; every mapping decision downstream depends on it.
Do we have to replace any systems to make them work together?+
Usually not, and it is worth resisting the reflex. Departments generally chose systems that fit their jobs; the failure is between the systems, not inside them, and integration attacks the between at a fraction of migration’s cost and risk. The exceptions are real but narrow: a system so closed it exposes no usable interface, or two systems whose overlap is so large you are paying to keep them in sync forever. In those cases consolidation belongs on the table. But the burden of proof sits on replacement — a working system that merely does not talk yet is a candidate for a connection, not a funeral.
In this cluster
Untangle it with us
An assessment that maps your systems and their copies of each record, names the ownership decisions waiting to be made, and identifies the one interface whose fixing would be felt soonest.
Start the conversation
