# Data Science Delivery Pack

Skill: `delivering-data-science`

[Word report](report.en.docx) · [Excel register](register.en.xlsx)

## Record details

| ID | Field | Project value |
| --- | --- | --- |
| `artifact_id` | Artifact ID | |
| `project` | Project | |
| `skill` | Skill | |
| `version` | Version | |
| `date` | Date | |
| `status` | Status | |
| `author` | Author | |
| `owner` | Owner | |
| `approver` | Approver | |
| `purpose` | Purpose | |
| `business_outcome` | Business outcome | |
| `scope` | Scope and exclusions | |
| `source_inputs` | Inputs and sources | |
| `assumptions` | Assumptions and constraints | |
| `decisions` | Decisions | |
| `rationale` | Rationale | |
| `risks` | Risks | |
| `exceptions` | Exceptions | |
| `open_items` | Open items | |
| `due_date` | Due date | |
| `acceptance_criteria` | Acceptance criteria | |
| `evidence_location` | Evidence location | |
| `reviewer_decision` | Reviewer decision | |
| `next_gate` | Next stage gate | |

## Domain analysis and decisions

### Business hypothesis and decision boundary (`hypothesis`)

Define the hypothesis, baseline, success criteria, and decisions not delegated to automation.

> Enter project evidence, conclusion, and owner.


### Data features and labels (`data_features`)

Fix prediction time, data purpose, feature availability, label delay, and leakage checks.

> Enter project evidence, conclusion, and owner.


### Experiment and evaluation protocol (`evaluation`)

Set temporal split, independent test, group errors, costs, and comparator metrics.

> Enter project evidence, conclusion, and owner.


### Reproducibility evidence (`reproducibility`)

Record data version, code, environment, randomness, and evaluation artifacts.

> Enter project evidence, conclusion, and owner.


### Model card and limitations (`model_card`)

State intended use, performance, known failures, affected groups, and human fallback.

> Enter project evidence, conclusion, and owner.


### Production service and rollout (`production`)

Define shadow or canary gates, human process, observability, and release approval.

> Enter project evidence, conclusion, and owner.


### Monitoring retraining and change (`monitoring`)

Set drift, effectiveness, bias, retraining triggers, and independent review.

> Enter project evidence, conclusion, and owner.


### Rollback and retirement (`rollback`)

Define stop thresholds, rollback, data retention, and retirement ownership.

> Enter project evidence, conclusion, and owner.


## Working registers

### Data Features (`features`)

Record feature sources, availability, and use constraints.

| Feature (`feature`) | Definition (`definition`) | Source (`source`) | Rule (`rule`) | Owner (`owner`) | Evidence (`evidence`) |
| --- | --- | --- | --- | --- | --- |
|   |   |   |   |   |   |

### Validation Monitoring (`validation`)

Track independent validation, drift, and stop actions.

| Metric (`metric`) | Validation (`validation`) | Drift threshold (`drift_threshold`) | Test result (`test_result`) | Owner (`owner`) | Evidence (`evidence`) | Review date (`review_date`) |
| --- | --- | --- | --- | --- | --- | --- |
|   |   |   |   |   |   |   |