SYNTHETIC DATA -- 100% machine generated. Granite Auto Group and every store, employee role and transaction referenced by this project are fictional. No real customer, employee, dealership or vendor data is present, and no record here describes a real person or a real business.
Granite Auto Group is a fictional dealer group. Every store, employee role and transaction in this export is machine generated.
The export carries the governed KPIs implemented to date. Lead-source quality, campaign cost and employee performance are modelled; no dataset here stands in for anything that is not.
management-actions is a DERIVED artifact. It reads no view: it is produced by evaluating config/dashboard/action_rules.yaml against the datasets in this same export, so every action can be recomputed by hand from files in the repository. Its rows are review prompts, not findings, recommendations of business action, or evidence of real-world conditions, and the queue is stateless -- regenerated with each dataset version, holding no history, acknowledgement, assignment or completion. No language model, learned model or scoring heuristic takes any part in producing it, and every threshold the rule file owns is a project default for a fictional dealer group.
The action register retains every proposed rule identifier, including those that do not fire. A disabled identifier carries the audited reason it is disabled: the project holds no such evidence, the evidence exists at a different grain, or the condition cannot survive into a valid export and belongs to the validation layer rather than to a management queue.
The GL control accounts are SYNTHETIC and there are three of them. They are a selected inventory control catalogue, not a chart of accounts, and no real dealer group's account numbering was consulted. ARPI models no journal entry, journal line, posting batch, trial balance, period close or financial statement, so nothing in this export is or supports a financial-statement assertion.
Both sides of the inventory reconciliation are generated from one governed model. It is not agreement between two independent systems, and the development dataset contains deliberately planted controlled variance scenarios so that all four comparison states are exercised. A nonzero variance is a position to investigate and is not evidence of an accounting error.
Inventory control balances are semi-additive. They add across stores and control accounts on ONE comparison date and never across dates; a period figure is the last comparable date within it. A missing side is published as null, never as zero, because a balance that does not exist is not a balance of zero.
market_price_estimate is a SYNTHETIC estimate generated for this fictional dataset. No auction result, guidebook, licensed benchmark or observed transaction is consulted anywhere in this project. price_to_market_ratio describes the asking price against it and is null where no estimate exists; neither is evidence that a price is right or wrong, and neither supports a repricing recommendation.
Price movement is derived from consecutive month-end snapshots of the same unit. ARPI holds no price-history fact and models no manager decision, pricing strategy or repricing action, so an observed decrease is an observation and nothing more.
Floorplan principal is liability context carried alongside a unit. It is not part of book value, is never netted against it, and ARPI publishes no net inventory position and models no floorplan interest, curtailment or carrying cost.
Manufacturer incentives, holdback and floorplan credits are excluded from front gross, so new-vehicle front gross is understated by construction. That is a modelling boundary, not a finding.
Ratios and rates are exported unrounded at the scale the reporting view produced. display_precision states how the console should render them; the export never discards the exact value.
Median, percentile, days-supply and inventory-turn figures are not additive. No group-level total is published for them, because a group median is not the average of store medians. Their evidence is row-level equality with the source view.
logical_run_key is null: ADR-0010's logical run key is recorded in the audit layer's pipeline-run table, the reporting layer does not publish it, and the exporter may not read that schema. It is left null rather than guessed.
Power BI real-engine validation remains pending on both ADR-0008 paths. Nothing in this export validates the semantic model, and no artifact here may be cited as Gate 2 evidence.
aged_threshold_days and the age-bucket boundaries are labelled project defaults, never industry benchmarks.
