Integrations
What to ask your IT team before integrating
Five questions that save three months of meetings and an integration nobody uses.
Most integrations that fail do not fail for technical reasons: they fail because nobody agreed beforehand what gets synced, in which direction, and who wins when the two systems disagree. That is settled in a half-hour conversation, not in code.
The five questions
- 1
Which data exactly, and in which direction?
"Sync suppliers" is not an answer. "Tax ID, name and document status, from here to the ERP" is.
- 2
Who wins if both change?
The question almost nobody asks and the one behind every later mess.
- 3
How often?
Instantly, hourly or nightly. It changes the work and the cost.
- 4
What happens when it fails?
Because one day it will. Who finds out and what they do meanwhile.
- 5
Who maintains it a year from now?
If the answer is "whoever built it" and that person could leave, you have an open problem.
Important
The second is decisive. Without a clear rule on who wins, the first time the systems disagree someone will fix it by hand in both places, and from then on nobody trusts either.
What to bring to that meeting
| Bring | What for |
|---|---|
| The concrete case, not "integrate" | An integration without a use case never finishes |
| Who suffers the manual work today | They will know whether the result is any good |
| How much time is lost now | The only way to know whether it pays |
| And the option of not integrating | Sometimes a monthly file export is enough |
Watch out
The last row is not rhetorical. Many integrations are proposed to save work that happens twice a month; maintenance costs more than that work. Check before you start, not after.
The three approaches, by effort
Worth knowing
Starting with the first and seeing whether it suffices is almost always the right call. You can extend later; unwinding a badly designed two-way sync is far more expensive.
›Do we need a developer?
For the first, barely. For the third, yes and on an ongoing basis.
›How long does it take?
The technical part, days. Agreeing what gets synced, as long as you take to decide.
›What if we have no IT team?
Then manual export and import is probably the right answer.
A real case
The situation
A company wants two-way supplier sync with its ERP.
What you do
- Answers the five questions and finds it only needs document status flowing to the ERP
- Sets up one-way notifications
What you get
It ships in days instead of months, and maintenance does not hinge on one person.
The situation
Integration is requested without knowing which fields are needed.
What you do
- Brings the specific flow and the field list
- States what triggers what and how often
What you get
IT estimates on data rather than on an intention.
The situation
IT asks about the upgrade calendar and nobody has checked it.
What you do
- Checks when each system changes before starting
What you get
The integration is not born obsolete.
The situation
It is built and nobody knows who maintains it.
What you do
- Names an owner and a deputy from the start
What you get
The integration is not orphaned.
The situation
There is no test environment and testing happens in production.
What you do
- Asks for a separate environment for testing
What you get
Go-live does not pollute real data.
The situation
Who does what is agreed verbally.
What you do
- Writes the split down before starting
What you get
Later doubts are settled by reading.
This article answers
- preparing an integration with our erp
- what do i need to integrate two systems
- meeting it before an integration
- integration readiness checklist