Protected ERP Analytics Pilot: Real-Derived Signals + 30 Analyst Tasks

About this data

Evaluate operational analytics and analyst agents with a compact, documented real-derived pilot. Includes 188 protected observations across three relative 28-day operational series, six aggregate first-receipt timing profiles, and 30 authored, machine-checkable analytic tasks. CSV and Parquet data, answer-free JSONL inputs, reference evaluation cases, four local forecasting baselines, a standard-library agent evaluator, data dictionary, transformation/privacy reports and reproducible integrity checks are included. Examples run locally without a paid AI API. The target is normalized recorded outflow workload, not physical inventory, market demand, revenue or cash flow. The reference agent example is scripted, not evidence of learned-agent performance. Internal exploratory baseline error reductions were 0%, 25.98% and 13.89% across the three cohorts; the smaller improvement is statistically uncertain. There is no external performance guarantee. One source and small coverage. Payment profiles are conditional on six retained, marked-paid groups and describe first linked receipts, not full settlement. Roughly 15% source receipt-link coverage limits interpretation. Original identities, exact dates, amounts, products and raw transactions are excluded. Aggregation, perturbation and content scans do not certify anonymity; organization-pattern linkage remains unquantified. USD 250 one-time license for this fixed edition and one purchasing entity. Internal commercial analytics, research, ML training and evaluation allowed. Non-reconstructive derived models may be commercialized. No dataset resale, redistribution, raw publication, sublicensing or re-identification. Future editions, custom consulting and support services are not included. See custom reuse terms. An evaluation preview is prepared and available on request. Review the scope and limitations before purchasing. CUSTOM COMMERCIAL LICENSE — FULL TERMS Protected ERP Analytics Pilot v0.2 — Commercial Internal-Use License Licensor: Alessandro Barrios. This text is proposed for the DataBazaar listing; it takes effect for a purchaser only upon an accepted paid order identifying the purchaser, this edition and these terms. No rights arise from this local draft. The paid license is non-exclusive, worldwide and non-transferable, for use by one purchasing individual or legal entity and its authorized personnel. It permits internal commercial analytics, research, software development, machine learning training and evaluation. There is no limit on internal experiments. Rights continue unless terminated for material breach, subject to mandatory law. The USD 250 price covers this fixed edition only; future editions, consulting, customization and support services are not included. Ownership is not transferred. No resale, redistribution, sublicensing of the dataset, raw publication, public hosting, or sharing with another legal entity. No attempted re-identification, linkage to the original organization or persons, credential recovery or unlawful surveillance. Preserve provenance notices. Do not present these aggregates as raw transactions, physical inventory, company-wide financial records or verified fraud labels. Contractors may access only on the purchasing entity's behalf, under equivalent restrictions; that entity remains responsible for their use. Non-reconstructive derived models and summary statistics may be commercialized. They must not expose source identities, permit substantial dataset reconstruction, or function as a dataset redistribution service. Do not publish substantial portions of the licensed dataset as examples or model outputs. Data is aggregated and perturbed from one real operational source. Direct identifiers, calendar dates, original amounts and raw records are excluded. Temporal organization linkage remains unquantified. This is not formal differential privacy, anonymity certification or an external utility guarantee. It is a small evaluation pilot: 188 observations in three relative 28-day series, six retained timing profiles and 30 authored analytic tasks. Timing profiles describe first linked receipts, not full settlement or business-wide behavior. Provided as-is to the extent permitted by applicable law. No promise of fitness, external prediction accuracy, industry representativeness or privacy certification. Nothing limits mandatory, non-excludable legal rights or overrides applicable marketplace dispute/refund procedures. Source database ownership is not transferred. For material breach, the Licensor may terminate the license by notice; cease dataset use and delete copies subject to mandatory retention. Compliant, non-reconstructive models/statistics may be retained unless mandatory law requires otherwise. No separate governing law or arbitration clause is imposed here.

Schema

Schema not available

Sample Data