Layered KPI framework for the Prime Products AI transformation — definitions, baselines, M6/M12 targets, owners, and Mission Control display.
KPI & Success Measurement Framework
How the transformation proves it worked. Targets are proposals to ratify at the M3 steering after baselines land; all baselines are TBD — captured in discovery (WBS 1.12) unless noted. Displayed-on pages refer to Mission Control.
Measurement principles
- Baseline first. Every KPI gets a measured (or honestly estimated) baseline during discovery M1–M3. No baseline → no target → not a KPI.
- Measure monthly, review at steering; company-level KPIs also in the board pack.
- Automate collection. Platform telemetry (WBS 5.6), agent run metrics (9.5), and KB health checks (6.6) feed Mission Control directly. Manually collected KPIs are limited to what SoftOne/M365 can’t emit, and each needs a named collector.
- Targets in two horizons: M6 (platform GA + pilots) and M12 (program end). M6 targets are deliberately modest — the platform barely exists before M6.
- Counter-metric discipline. Every efficiency KPI is paired with a quality guard (e.g. quote turnaround ↓ must not push quote error rate ↑).
- No target-gaming. KPI definitions are frozen once ratified; changes require a steering decision recorded in the minutes.
Column key: Baseline = TBD (discovery) unless stated. Owner = role accountable for the number, not for collecting it.
1. Company-level KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Revenue per employee | Annual revenue ÷ headcount (~€23.7M / ~109 ≈ €217k today) | SoftOne + HR | ≈€217k (2025, to confirm) | n/a (annual) | +3–5% run-rate | CEO | Executive dashboard |
| Net margin | Net profit ÷ revenue (thin ~1.5% today — the program’s core financial lever) | Finance | ≈1.5% (to confirm) | n/a | +0.3–0.5 pp run-rate contribution | CEO | Executive dashboard |
| Quote turnaround time | Median hours from RFQ received → quote sent | SoftOne + email timestamps (manual sample until integrated) | TBD | −20% in pilot depts | −40% company-wide | Sales head | Executive dashboard |
| Order-to-delivery lead time | Median days from confirmed order → delivered to vessel/customer | SoftOne | TBD | baseline held (no regression) | −15% | Logistics head | Executive dashboard |
| Tender/bid throughput | Bids submitted per month per bid-team FTE | Sales records | TBD | +10% | +30% | Sales head | Executive dashboard |
| Working-capital days (DSO) | Days sales outstanding | Finance | TBD | baseline held | −5 days | Finance head | Executive dashboard |
2. Department-level KPIs
One KPI set per department, defined in the 04-departments one-pagers (WBS 2.8) and rolled up here. Representative examples:
| Department | Lead KPI (definition in dept file) | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Sales | Quotes issued per salesperson per week | TBD | +10% (pilot) | +25% | Dept head | Dept heatmap |
| Customer service | First-response time to customer inquiry | TBD | −30% (pilot) | −50% | Dept head | Dept heatmap |
| Procurement | Supplier-quote comparison cycle time | TBD | — | −30% | Dept head | Dept heatmap |
| Warehouse | Picking/document error rate per 1000 lines | TBD | — | −25% | Dept head | Dept heatmap |
| Compliance/quality | Audit-prep effort (person-days per audit) | TBD | −25% (if pilot) | −50% | Dept head | Dept heatmap |
| Finance/accounting | Month-end close duration (working days) | TBD | — | −2 days | Dept head | Dept heatmap |
| Reporting/BI | Ad-hoc report delivery time | TBD | −30% | −60% | Dept head | Dept heatmap |
Remaining departments (marketing, HR, logistics, product/technical support, IT, administration, executive) carry their lead KPIs in their department files; all appear on the department heatmap.
3. Process-level KPIs
Attached to the mapped processes (WBS 2.2–2.4); measured for processes touched by a use case.
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Process cycle time | Elapsed time start→end per mapped process instance | SoftOne timestamps / sampling | TBD per process | −15% (piloted processes) | −30% (automated processes) | Process owner | Process scorecards |
| Touches per transaction | Human interventions per order/quote/document | Process sampling | TBD | −1 touch (piloted) | −30% | Process owner | Process scorecards |
| Rework loops | % of instances returned for correction | Sampling + agent logs | TBD | measured | −25% | Process owner | Process scorecards |
4. AI adoption KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Platform WAU | Unique users active in trailing 7 days | Telemetry | 0 | ≥25 (pilot depts) | ≥40 of ~50 target users | TPL lead → IT | Adoption dashboard |
| WAU/MAU stickiness | WAU ÷ MAU | Telemetry | — | ≥0.5 | ≥0.7 | TPL lead → IT | Adoption dashboard |
| % employees active weekly | WAU ÷ 109 | Telemetry + HR | 0 | ≥20% | ≥65% | CEO | Adoption dashboard |
| Queries per user per day | Total queries ÷ DAU | Telemetry | — | ≥3 | ≥8 | TPL lead → IT | Adoption dashboard |
| Assistant CSAT | Avg in-tool rating (thumbs → 1–5 pulse) | In-tool feedback + pulse survey | — | ≥3.8/5 | ≥4.2/5 | TPL lead | Adoption dashboard |
| Departments with weekly active use | Depts where WAU ≥50% of staff | Telemetry | 0 | 2–3 (pilots) | ≥13 of 15 | Steering | Dept heatmap |
5. Knowledge base KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Docs count | Knowledge docs with valid frontmatter | KB health check (6.6) | ~0 | ≥150 | ≥500 | TPL lead | KB health |
| SOP coverage | % of mapped processes with a current SOP doc | KB check vs process inventory | TBD (low) | ≥60% pilot depts | ≥80% all depts | Dept champions | KB health |
| Freshness | % docs reviewed/updated within their review cycle | Frontmatter dates | — | ≥70% | ≥85% | Dept champions | KB health |
| Orphan rate | % docs with no inbound links | Link graph check | — | ≤15% | ≤8% | TPL lead | KB health |
| Retrieval hit rate | % RAG queries where a retrieved chunk is used in a rated-useful answer | RAG eval + feedback (5.4, 6.7) | — | ≥70% | ≥85% | TPL lead | KB health |
6. Automation KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Registered agents in production | Count in ai-agent-registry.md | Registry | 0 | 0–2 | ≥10 | TPL lead → IT | Automation showcase |
| Agent runs per month | Successful runs across all agents | Agent telemetry (9.5) | 0 | — | ≥1,000 | TPL lead → IT | Automation showcase |
| Hours saved per month | Σ (runs × validated minutes-saved per run type) | Agent telemetry + time-study validation | 0 | measured for pilots | ≥400 h/month | Steering | Hours-saved counter |
| Straight-through rate | % agent runs completing with no human correction | Agent telemetry | — | measured | ≥70% avg | TPL lead → IT | Automation showcase |
| Human-override rate | % runs where output rejected/heavily edited | Agent telemetry | — | measured | ≤15% | TPL lead → IT | Automation showcase |
7. Infrastructure KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Platform uptime | % availability in business hours (07:00–19:00 EET) | Monitoring | — | ≥99.0% | ≥99.5% | Prime Products IT | Platform status |
| p95 chat latency | 95th-percentile time-to-first-token | Telemetry | — | ≤2.5 s | ≤2.0 s | Prime Products IT | Platform status |
| GPU utilization (business hours) | Mean GPU busy % | Monitoring | — | 20–50% (headroom) | 40–70% | Prime Products IT | Platform status |
| Cost per query | (Amortized CapEx + OpEx)/month ÷ queries/month, € [placeholder model] | Finance + telemetry | — | measured | declining month-over-month | Finance head | Platform status |
| Incident count (Sev1/Sev2) | Production incidents per month | IT incident log | — | ≤2 | ≤1 | Prime Products IT | Platform status |
8. Training KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| People trained | Staff completing their audience track (change plan) | Training records | 0 | ≥25 (champions + pilots) | ≥100 of ~109 | TPL lead | Training hub |
| Certification rate | % trained staff passing internal certification (10.7) | Assessment records | 0 | — | ≥70% | TPL lead | Training hub |
| Prompt-skill assessment | Median score on standardized prompt-task rubric | Assessment (10.7) | — | measured for pilots | ≥75/100 median | TPL lead | Training hub |
| Champions active | Champions attending ≥3 of 4 monthly guild meetings | Guild attendance | 0 | 15 nominated | ≥12 active | CEO | Training hub |
9. Value: ROI, productivity, time saved
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Program ROI indicator | (Annualized hours-saved value + margin-effect estimate) ÷ program cost €[TBD] | Finance model, inputs from §6 | — | model built | ≥1.0× run-rate, path to 2–3× yr-2 | Steering | Executive dashboard |
| Time saved per employee per week | Self-reported + telemetry-derived hours | Pulse survey + §6 | 0 | ≥1 h (pilot users) | ≥3 h (active users) | Steering | Executive dashboard |
| Output per FTE in AI-assisted tasks | Task-specific throughput (quotes, replies, reports) per FTE | Dept KPIs (§2) | TBD | +10% pilots | +20–30% assisted tasks | Dept heads | Dept heatmap |
| Overtime hours in assisted departments | Monthly overtime, HR records | HR | TBD | — | −15% | HR head | Executive dashboard |
10. Quality KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Quote/document error rate | Errors found post-issue per 100 quotes/docs (counter-metric to speed KPIs) | Sampling + CS complaints | TBD | no regression | −20% | Dept heads | Process scorecards |
| Rework rate | % deliverables redone after review | Sampling | TBD | no regression | −25% | Dept heads | Process scorecards |
| Compliance findings | ISO 22000 / bonded-warehouse audit non-conformities per audit | Compliance records | TBD | no regression | −30% | Compliance head | Process scorecards |
| AI-output defect rate | % AI-assisted outputs failing human verification (verification duty per usage policy) | Feedback + sampling | — | measured | ≤5% on production use cases | TPL lead | Adoption dashboard |
11. Customer service KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| First-response time | Median time to first substantive reply to customer inquiry | Email/Teams timestamps (manual sample → integrated) | TBD | −30% (if pilot) | −50% | CS head | Dept heatmap |
| Resolution time | Median time inquiry → resolved | CS records | TBD | −15% | −35% | CS head | Dept heatmap |
| Same-day quote rate | % RFQs answered with a quote same business day | SoftOne + sampling | TBD | +15 pp (pilot) | ≥80% | Sales head | Executive dashboard |
| Customer complaints | Complaints per 100 orders | CS log | TBD | no regression | −20% | CS head | Dept heatmap |
12. Management visibility KPIs
| KPI | Definition / formula | Source | Baseline | Target M6 | Target M12 | Owner | Mission Control page |
|---|
| Report latency | Days from period end → management report available | Reporting/BI records | TBD | −30% | same-day/next-day for standard pack | Reporting/BI head | Executive dashboard |
| Ad-hoc question turnaround | Median time from management data question → answered | Sampling | TBD | −40% | minutes (self-serve via assistant) | Reporting/BI head | Executive dashboard |
| Decision lead time | Days from issue raised at steering/management → decision recorded | Meeting minutes | TBD | measured | −30% | CEO | Executive dashboard |
| KPI automation rate | % of KPIs in this framework auto-collected (principle 3) | Mission Control config | 0 | ≥50% | ≥80% | TPL lead | KB health |