Why your systems do not talk to each other

Almost every business past a certain size has one: a person, or several, who exist partly to move information from one system into another. An order arrives in one place and is re-entered somewhere else. A record is updated twice, and the two copies drift.

It is worth naming what that is. It is not a staffing problem or a discipline problem. It is two systems that were bought separately and never introduced.

Three reasons it persists

• Nobody owns it. Each system has an owner; the gap between them does not. Work that falls between two owners does not get prioritised.
• The cost is invisible. Nobody invoices for the twenty minutes a day. It shows up as headcount, not as a line item, so it never competes for budget.
• Someone quoted a platform. The first answer to “our systems do not talk” is often a large integration platform with a licence to match. That is sometimes right and frequently enormous overkill for two systems and one direction of data.

What connecting two systems actually involves

Less than the platform quote suggests, and more than a script. The work is usually four things:

• Agreeing what is true. When the two systems disagree about a customer’s address, which one wins? This is a business decision and it has to be made before any code.
• Getting the data out and in. Modern products have an API. Older ones have a file drop or a database. Both are workable; the second is slower to build against.
• Handling failure. What happens when the other end is down, or rejects a record? A sync with no error handling silently loses things, which is worse than not having it.
• Making it visible. Somebody has to be able to see that it ran, and what it did, without asking an engineer.

What it does not require

It does not require replacing either system. It does not usually require a licensed integration platform. And it does not require the two vendors to cooperate — if both expose their data, the connection can be built without either of them being involved.

It also does not require doing everything at once. One direction, one record type, running reliably, is worth more than a complete two-way sync that is still in testing.

How to size it before you ask anyone

Count the number of record types that move, the directions they move in, and how quickly they need to arrive. Three record types moving one way, hourly, is a small piece of work. Everything moving both ways in real time is a large one. Most businesses need the first and are quoted for the second.

Common questions