Four systems, four versions of the truth
Finance says one number, operations says another, and the meeting turns into an argument about whose export is right.
What it looks like from the inside
This is rarely a dashboard problem. It is that nobody agreed what a customer is, the nightly job fails quietly on Sundays, and one of the four sources is a spreadsheet that somebody edits by hand. Reporting on top of that produces confident charts that disagree.
Roughly half the AI projects we are asked about turn out to be this. It is worth knowing early, because fixing it is usually the larger half of the work and the part that makes everything after it possible.
- Two reports on the same thing, and nobody trusts either
- A job failed on Sunday and you found out on Wednesday
- Definitions live in people's heads: what counts as active, as a customer, as done
- One critical source is a spreadsheet one person maintains
What we do about it
Map what you have
Where each number comes from, who owns it, and which definitions disagree. This is the part that produces the uncomfortable meeting, early, when it is cheap.
Build the pipeline with the checks in it
Ingestion, transformation and quality checks that hold a bad batch back instead of publishing it. A source that breaks does not take the others down with it.
Say how fresh the answer is
Every dashboard states when its data last arrived, so a stale number cannot pass for a current one.
Alert the people who can act
Monitoring goes in from day one, pointed at the team that owns the fix rather than at a mailbox nobody reads.
What you end up with: a warehouse with agreed definitions, pipelines that fail safely, and dashboards that say how fresh they are
The services this draws on
Other weeks that are not working
Does this sound like you?
Send a paragraph about how it shows up in your week. We reply within two working days with questions, a rough shape, and an honest answer on whether we are the right people for it.
Start a project