Why your business systems do not talk to each other
Key takeaways
Every growing business ends up with systems that each know half the picture. Why it happens, why it is nobody's fault, what it really costs, and why connecting what you have usually beats replacing it.
Almost every growing business arrives at the same place. The shop knows about orders, the accounts package knows about money, the CRM knows about customers, and the stock system knows about what is on the shelf. Each one is good at its job. None of them knows what the others know.
So people become the connection. They retype, they reconcile, they chase a number from one screen to another, and nobody ever quite sees the whole picture, because half of it lives in one system and half in another.
Here is why it happens, why it is nobody's fault, and what to do about it.
Why it happens
Each system was bought to solve one problem, at one time. The accounts package came first because the accountant asked for it. The shop came when you started selling online. The CRM came when the customer list outgrew a spreadsheet. Every purchase was sensible, and none of them was ever chosen with the others in mind.
Joining them was nobody's job. The shop supplier sells shops. The accounts supplier sells accounts. Each will tell you, accurately, that their product connects to things. Neither is responsible for making it connect to yours.
The gap was small at first. Copying ten orders a week by hand is nothing. Copying ten a day is an hour. By the time it is a problem, it has become a process, and processes do not get questioned the way problems do.
None of that is a failure of management. It is what happens to every business that grows a piece at a time, which is all of them.
What it actually costs
The obvious cost is time, and it is the smallest one.
The bigger cost is errors. Every time a person retypes a record there is a chance of a slip, and slips in money and stock are expensive in ways that do not show up as a line on anything.
Bigger still is not seeing clearly. If you cannot answer "which customers bought this month and have not paid" without somebody spending an afternoon cross-referencing two systems, you will mostly not ask the question. Decisions get made on whichever half of the picture is easiest to look at.
The two ways to fix it
Replace several systems with one. Some businesses genuinely should. If everything you use is old, unsupported, or a poor fit, moving to one platform that does most of it can be right. But it is the expensive option, it is slow, and it forces you to change how you work to match how the new system expects you to.
Connect what you already have. Usually cheaper, usually faster, and it keeps the tools your staff already know. Each connection between two systems is a separate, small job. You can do the one that hurts most first, see it working, and decide about the next one later.
Most of the time the second is the better answer. The systems you have are mostly fine. They have never been introduced.
Where to start
Pick the one connection that costs the most. Not the most technically interesting one, the one where the most time is lost or the most mistakes happen. In most businesses it is one of three: online orders into the accounts package, enquiries into the CRM, or orders into the stock system.
Measure it for a fortnight: how long, how often it goes wrong, what a mistake costs. That gives you a number to judge any quote against, and it is the first thing a good supplier will ask you for anyway.
What to be wary of
A proposal that starts with replacing everything. It might be right, but it should be argued for, not assumed.
A connection that runs under the supplier's own accounts. If the credentials that let it read your systems belong to them, then leaving them means it stops. You own your accounts, you licence the code, and you can leave whenever you like.
And anything that cannot tell you when it has failed. A connection that silently stops is worse than none, because you have stopped checking.
What we do
Connecting systems is ordinary work here. Each connection is graded and priced on its own, so you can do one, see it working, and stop there if that is enough.
You describe which systems and what has to move between them. You get a price in writing before anything starts, and nothing begins until you approve it.
Nothing we have not already seen. Thirty years of this means most of these connections are familiar shapes, and the price reflects that.
The price changes with the size of the job. The standard never does.