Representative demonstration
Invented manufacturing records demonstrate the workflow and output format without using customer information.
LEFTFOUNDING MANUFACTURER PROGRAMMEProof Sprint · £2,950–£4,500
RESULTS & EVIDENCE
Evidence you can inspect.
Confidence you can build.
Explore the product checks, representative scenarios, quality experience and measurement approach behind RepeatProof. Each form of evidence is clearly identified so you can assess the product on its merits.
PRODUCT EVALUATION6 representative scenariosRELEASE CHECKS42 required checksQUALITY EXPERIENCE25+ years in manufacturing qualityMEASURED OUTCOMESPublished with customer approval
EVIDENCE FRAMEWORK
Four forms of evidence.
One clear standard.
Representative demonstrations, versioned product checks, manufacturing-quality experience and customer-approved outcomes are presented with the context needed to understand each one.
Versioned product checks
6 labelled evaluation scenarios test expected signals, evidence traceability, decision boundaries and deterministic repeat.
Inspect the benchmark ↗Manufacturing-quality experience
Founder experience shapes the operating method, delivery approach and practical quality boundaries.
Review founder experience ↗Customer-approved outcomes
Published with the agreed baseline, method, sample, limitations and recorded customer approval.
Open the outcome register ↗CONCISE DATA FLOW
A bounded calculation path.
A named decision path.
Exact providers, regions, retention and deployment boundaries are agreed for each customer. This summary describes the application pattern, not a universal customer configuration.
Approved historical export
The customer data owner provides only the records and fields in the signed Proof Sprint boundary.
Tenant-isolated workspace
Access is restricted to authorised members. Raw identifiers and files remain inside the controlled workspace.
Versioned calculation engine
Normalisation, retrieval, recurrence, measurement and cost rules produce source-linked evidence with confidence and limitations.
Named human review
The reviewer accepts, edits or rejects the prepared output. The QMS and authorised people retain the official decision.
ACCESS MATRIX
Least privilege
made visible.
Roles are confirmed in the engagement plan. The matrix below is the default Proof Sprint boundary, not a substitute for the customer's approved access design.
| Role | Approved records | Prepared analysis | Quality decision | Configuration |
|---|---|---|---|---|
| Reviewer / Quality Manager | Read scoped | Review, edit, reject | Authorised in customer process | No |
| Data owner | Prepare approved export | Inspect mapping | No | Data boundary only |
| RepeatProof delivery lead | Minimum scoped access | Configure and support | No | Proof Sprint scope only |
| RepeatProof calculation engine | Process scoped fields | Prepare source-linked result | Never | Versioned rules only |
Available due-diligence evidence
Prospective customers can request the architecture and data-flow summary, role/access matrix, retention and deletion schedule, model-processing description, release and incident runbooks, subprocessor list and security questionnaire response. Certification or penetration-test evidence is supplied only when it exists.
PUBLIC OUTCOME TEMPLATE
The proof asset
every approved result must complete.
This is a publication template, not a result. A customer-approved outcome will fill every field and state the measurement method and limitations beside the numbers.
| Required field | What must be disclosed |
|---|---|
| Dataset and baseline | Sources, date range, record count, exclusions, current preparation time and agreed start date. |
| Labelled benchmark | How the known-relevant set was created, by whom, and any unresolved judgements. |
| Precision | Reviewer-accepted related records divided by all displayed records. |
| Recall | Known relevant records retrieved divided by all records in the known-relevant set. |
| Top-K recall | Recall at the first 5, 10 and 20 displayed records. |
| False-negative rate | Known relevant records missed, reported as a count and percentage. |
| Reviewer agreement | Raw decision agreement on independently reviewed records, plus disagreement reasons. |
| Preparation-time result | Comparable median assisted time using the same start and stop definition. |
| Operational findings | Accept, edit and reject counts, recurring correction patterns and observed workflow constraints. |
| Limitations | Missing fields, selection effects, short observation period and constraints on generalisation. |
Set the measurement method
before the result.
The Proof Sprint plan fixes preparation time, precision, recall, Top-K recall, false-negative rate and reviewer-agreement definitions before delivery.