Service

Contested Metric Diagnostic

Two official numbers. Both claim truth. We trace one metric from the source through the warehouse into the report and leave a written ruling plus evidence SQL — typically in about ten business days.

Start a diagnostic

One metric

The number both sides claim

Two artifacts

The reports or extracts in dispute

About ten business days

A fixed, short statement of work

Written ruling

Which number is valid for which decision

Evidence SQL

Queries your team can re-run

Stack

SnowflakedbtPythonSQLAWSAirflowTableau / Power BI

The work

How we help

When the jobs are green and both dashboards refreshed, the problem is not a failed pipeline. It is the level of detail — an order total versus the line items — a filter, a copied rule, or a delete the load dropped.

The diagnostic is a fixed, short statement of work. Building the fix is a separate engagement if you want it.

When this is the engagement

Both dashboards refreshed, and nobody will yield — close, commissions, or an agent is about to pick a side.

Contracted and invoiced by Havilah Technologies LLC. Delivery in your warehouse and your git.

01Take one metric and two artifacts — a report, an extract, a board pack, or an agent answer.

02Trace the number from the operational document through ingest, the warehouse, the mart, and the report.

03Issue a written ruling: which number is valid for which decision.

04Deliver evidence SQL your team can re-run.

05Rank the fixes, with a list of objects that must not be edited without a named approver.

Typical work

What an engagement looks like

Intake

The metric name, two artifacts, and access to the warehouse and the workbook.

Trace

Source, warehouse, and report — until the break is named.

Ruling

Which figure is valid for close, operations, or compensation.

Handoff

Evidence SQL and ranked changes. Implementation is a separate scope.

What you receive

Deliverables

A written ruling: which number is valid for which decision
Evidence SQL your team re-runs
Ranked changes with owners
A do-not-touch list for bonus formulas, GL maps, and load filters

Capabilities

In scope

One metric, traced from source to report
A written ruling for a named decision
Evidence SQL your team re-runs
Ranked fixes, not a silent rebuild

Method

Trace the number from source to report

Operational document

Invoice, job, contract — header vs line.

CDC / ingest

Deletes, late arrivals, what the pipeline actually kept.

Warehouse fact

Grain. The hop most fights actually live in.

Certified mart

The table both sides call source of truth.

BI calculation

Joins, filters, table calcs nobody remembers putting in.

Downstream

Bonus, branch scorecard, or the agent about to quote it.

01Ruling: which number is valid for which decision (close vs ops vs compensation).
02Break class: source, grain, mapping, late CDC, report logic, or a copied business rule.
03Evidence SQL — parameterized, they re-run.
04Ranked changes, hours, and owner.
05Do-not-touch list: bonus formulas, GL maps, CDC delete filters unless a named human approves.

Questions

How long is it?: Typically about ten business days for one metric and two artifacts. The scope is fixed and short.
What do we receive?: A written ruling and evidence SQL your team can re-run. A sample ruling is published only from a real engagement, with the client’s permission.
Does the fee include the rebuild?: No. The diagnostic names the break. Implementation is a separate written statement of work if you want it built.
What should we send?: The metric name, the two artifacts that disagree, and the decision or deadline that depends on the number.

Related services

Discuss Contested Metric Diagnostic.

Contested Metric Diagnostic is contracted under Havilah Technologies LLC. Tell us the stack, the systems, and the outcome.

Start a diagnostic