IncidentSEV2INC-2058

Test is right, Production is wrong

After Monday's release, Production shows nothing for May and last month's targets, while Test looks perfect. Same model, same report. Find out why.

Reported by: Jordan Reed, VP Sales
Opened: Wed 3 Jun 2026, 08:47
Needed by: Today
Priority: High
Time box: 3 h

Labels: productionreleasefinance

Previously at Northwind

The release from Sprint 06 was promoted through the deployment pipeline on Monday evening: Development → Test (checked by Nina) → Production.

Why the business cares Regional managers plan this week's targets from Production. With no May sales and last month's targets they will under-plan, and anyone comparing Production with Test will stop trusting both.

Brief

Jordan ReedVP Salesemail Wed 3 Jun, 08:47

May is blank??

The sales app shows nothing for May and the targets look like last month's. Nina says the Test version is fine. What happened on Monday?

Elena TorresBI Lead (your manager)chat Wed 3 Jun, 08:55

Sam deployed Test → Prod on Monday at 18:12 and refresh says Completed every day since. Please find out why Prod differs from Test, fix it properly (not by editing Prod by hand) and tell me how we stop it happening again.

Kenji MoriData Engineerchat Wed 3 Jun, 09:10

Nothing changed on the warehouse side. Prod DW loaded fine last night: max order date is 1 June. Oh, and I still need to drop that load-test copy I restored on the prod server in March, sorry.

What you are expected to do

Explain every difference between Test and Production in stage_comparison.csv, with the piece of configuration that causes it.
Fix Production through the pipeline (rules, bindings, credentials), not by editing the published model by hand.
Verify Production against the warehouse after the fix.
Propose preventive controls: what should a deployment check before and after it runs?

Deliverables

Root cause: each Test/Prod difference mapped to its cause
The configuration changes (rules, connection, credential owner)
Post-deployment checklist (automatable)
Runbook entry: how to diagnose "Test fine, Prod wrong" next time

Evidence

deployment_pipeline.json Stages, workspaces and deployment rules.
parameters_by_stage.csv Parameter values in each stage's semantic model after Monday's deployment.
sql_databases.csv
datasource_bindings.csv Gateway and cloud connections bound to each stage's model.
refresh_history_prod.csv
stage_comparison.csv A few checks Nina ran in Test and Production this morning.