Skip to main content
CodeCobra answers questions from your live data. The quality of the answer depends a lot on how you frame the question.

Questions that work well

“Why is this number what it is?” — the strongest use. Give it the number you’re looking at and where you’re looking at it.
“The dashboard says 312,000 revenue for March but the P&L says 298,000. Which is right and why do they differ?”
“How many, how much, which ones?” — counting and filtering across models that don’t have a ready-made report.
“Which customers have an open invoice more than 60 days old and an active subscription?”
“Reconcile these two things.” — anywhere two systems or two reports disagree.
“Compare our stock on hand against what the last physical count recorded, and list every product where they differ by more than five units.”
“Is this getting better or worse?” — trends, when you don’t have the report for it.
“Late deliveries per month for the last year, by warehouse.”

Give it the landmarks

You don’t need technical detail, but concrete anchors help:
  • Where you’re looking — “on the Sales dashboard”, “in the aged receivables report”
  • The period — “last month”, “Q1”, “since we went live in March”
  • The number you expected — “it should be about 400”
  • Anything you’ve ruled out — “it isn’t the tax, I checked”
Paste the number you’re disputing. “It says 12,431.88 and I think it should be 11,900” gives CodeCobra a target to reconcile against, which is much faster than “the total looks wrong”.

Reading the answer

A good answer has three parts: what the number actually is, why it differs from what you expected, and the records that demonstrate it. If you only get the first, ask for the rest — “show me the specific orders”.

When it can’t answer

Sometimes the data genuinely isn’t there — a field was never filled in, or the history predates your Odoo instance. CodeCobra will tell you that rather than produce a confident number it can’t support. That’s the intended behaviour; treat a hedged answer as information, not a failure.

Following up

The chat keeps its context, so you can drill in without starting over:
“OK — of those 43 late orders, how many were late because of the same supplier?”