Cost Model Steward | Yarken Help

Use this article to understand what the Cost Model Steward does, how it differs from the other specialists, and how it supports trust in the Yarken cost model.

Overview

Cost Model Steward focuses on Yarken’s governed semantic layer: mapping rules, coverage, data quality, split methods, and the audit trail. It explains why the model produces a figure, quantifies the effect of model issues, and proposes auditable structural fixes for human approval.

It owns the model and its quality. It does not provide business commentary about whether spend is good, bad, over budget, or commercially favourable.

What Cost Model Steward helps you do

Measure mapped and unmapped spend coverage.
Identify wrong-level, stale, duplicated, or incomplete mappings.
Quantify each model-quality issue in dollars and as a share of period spend.
Explain how a mapping or split rule affects reports.
Propose one auditable change for each identified issue.
Estimate restatement impact before a change is approved.
Review human and agent actions where audit data is available.
Maintain a clear human-approval boundary for model changes.

How the analysis works

The specialist states the reporting period and modelling progress.
It quantifies every finding in dollars and as a share of period spend.
It classifies the issue.
It proposes one change per issue, including the current behavior, proposed behavior, spend moved, and reports affected.
Every proposal remains a draft until a named human approver records approval.
It discloses the expected restatement effect before approval.
It reports open questions and limitations where rule or audit data is incomplete.

Issue classifications include unmapped, wrong-level, stale rule, split mismatch, duplicate master, ingestion failure, and unexplained.

Data sources and availability

Data source or tool	Purpose
report_spend_mapping_quality	Required source for unmapped spend, coverage, and dedicated model-quality metrics.
report_spend	Quantifies the spend affected by a gap or proposed rule change.
search_dimension_values	Resolves vendor, account, cost-centre, and other model names before a rule is proposed.
rule_inventory_read	Provides mapping rules and split methods across Technology Cost Pools, Technology Resource Towers, Solutions, assets, and Consumers.
audit_log_read	Shows who changed what and when for human and agent-proposed actions.

Complete model-health analysis requires report_spend_mapping_quality. Rule-level and audit findings require their corresponding read capabilities. The specialist does not infer an unmapped percentage from aggregate spend alone.

Output

Responses can include:

Coverage headline.
Model-quality issue table.
Dollar exposure and share of period spend.
Current and proposed rule behavior.
Spend moved and reports affected.
Restatement impact.
Required approver and open questions.
Data and analysis limitations.

Assigned skills

The finalized skill scope includes:

Model health review
Data quality assistant
Human change & override audit
Agent action audit

Audit skills require a durable audit trail and relevant approval state to be available.

Type	Cataloged workflows
Playbooks	Mapping exception handling
Automations	Mapping / DQ monthly check; Human override audit; Agent governance audit

Automations require the relevant data source, schedule, approval, and notification configuration. They must not generate false model-quality findings when a required source is unavailable.

How is Cost Model Steward different?

Cost Model Steward owns the structure and trustworthiness of the cost model. It explains mapping coverage, split rules, data-quality gaps, and the impact of changing the model.

IT Cost Analyst explains financial variance using the model’s outputs. Chargeback & Showback Analyst explains the consumer bill produced by allocation. Application & Service TCO Analyst explains the layered cost of a delivered application or service. Cost Model Steward steps in when the question is whether the underlying mapping or allocation structure can be trusted or should change.

Sample scenarios

Scenario 1: Assess whether the model can be trusted this period

A TBM administrator asks, “Can we trust the model this period?” The specialist reports mapped and unmapped spend coverage, ranks gaps by dollar exposure, classifies the main issues, and identifies reports that may be affected. If the dedicated mapping-quality source is unavailable, it states that the assessment is blocked rather than deriving a percentage from aggregate spend.

Scenario 2: Propose a mapping-rule correction

A new vendor is appearing as unmapped spend. The specialist resolves the vendor name, quantifies the affected spend, reviews the current rule inventory, and proposes one Technology Cost Pool or Technology Resource Tower assignment. The proposal includes current behavior, proposed behavior, spend moved, affected reports, and restatement impact. Nothing changes until a named human approver records approval.

Scope and hand-offs

The specialist handles semantic-model coverage, mapping rules, split methods, data quality, restatement impact, and auditability.

It hands business spend interpretation to the specialist that owns that domain, such as IT Cost Analyst, Cloud FinOps Analyst, or Chargeback & Showback Analyst.

It does not:

Answer general business-spend questions.
Infer model-quality metrics from an unsuitable source.
Commit a rule change without explicit recorded human approval.
Hide restatement impact.
Invent mappings, audit events, or approvers.

Next step

Return to Specialists to compare the default specialists.