Playbook A · conservative, tool-led
A · Scope & Automate
Classic risk-based testing done properly with today’s SAP-ecosystem tools and AI assistants. The lowest-risk option for customers who want a standard vendor stack and a predictable procurement path. Expect roughly a 2× reduction, not 10×.
2.1×
first deployment vs 4,820 PD baseline estimate
2.3×
mature / productised estimate
2.2× / 2.4×
ceiling at very large scale (first / mature) estimate
~13 months
QA calendar for a two-step IBM i path estimate
When to offer it
- Risk-averse customers, strong audit culture, preference for SAP-endorsed tooling (Cloud ALM, Tricentis as SAP Solution Extension).
- Small programs where an engine would not pay off (see sensitivity table).
- As a fallback or base layer under B or C.
Timeline
048121620242832364044485256
0 · Evidence-based scoping
1 · Generate & automate
2 · Unicode step validation
3 · S/4 mock cycles
4 · UAT, dress rehearsal, cutover
Calendar gains are smaller than effort gains: the technical steps (Unicode conversion, DMO runs, downtime tests) set the pace. Baseline QA calendar for the same path: ~14–16 months estimate.
Phases & activities
0 · Evidence-based scoping wk 0–6
- Readiness Check, simplification items, SCMON/SUSG + ST03N usage analysis
- Process mining (Signavio PI or Celonis) to identify live process variants
- Impact analysis (LiveCompare or Panaya) to rank processes by risk
- Agree scope cut-off and risk classes with process owners
1 · Generate & automate wk 4–18
- AI-assisted test case generation in Cloud ALM / ECT (from process models, existing docs)
- Automate top ~60% of in-scope regression (Tosca/Worksoft/Opkey-class)
- Set up DTV (or Kyano/ARC) report variants for FI/CO, MM, SD
- Build KID register (intended differences)
2 · Unicode step validation wk 14–22
- Run DTV/table diffs before vs after Unicode conversion
- Targeted tests on collation-sensitive reports, range selections and IFS file interfaces
- Fix-and-retest loop
3 · S/4 mock cycles wk 22–44
- Automated regression per mock, DTV diffs per mock
- AI duplicate-defect detection and triage support
- Coverage dashboards in Cloud ALM
4 · UAT, dress rehearsal, cutover wk 42–56
- Guided UAT on the reduced scope (Panaya / Cloud ALM / Original Software)
- Dress rehearsal with full DTV run in downtime
- Evidence pack and sign-off
Tools
- SAP Readiness Check, SCMON/SUSG, ST03N
- SAP Signavio Process Insights or Celonis
- Tricentis LiveCompare or Panaya
- SAP Cloud ALM + Test Management Assistant; SAP ECT by Tricentis / Tosca or Worksoft / Opkey / Avo
- SAP DTV (LTVF if restructuring) or SNP Kyano Validate / Natuvion ARC
Roles
- Test manager (customer or ecociel)
- Automation engineers (2–3)
- Functional leads per module (FI/CO, MM, SD, PP)
- Key users for UAT (~20 instead of ~30)
- Independent QA reviewer
Prerequisites
- ≥ 12 months of usage data including a year-end (start SCMON early)
- Cloud ALM tenant; tool licences
- Stable sandbox from the first conversion run
- Agreed risk classification and scope cut-off rules
Risks & mitigations
- Scope cut too aggressive → missed rare processes. Mitigation: keep year-end/quarter-end variants in scope regardless of frequency.
- Automation build eats savings. Mitigation: automate only processes that run in ≥3 cycles.
- Vendor claims don’t transfer to conversions. Mitigation: pilot on one module with measured before/after.
- DTV content gaps for custom reports. Mitigation: add custom ALV variants early.
Effort build-up per bucket estimate
Baseline = the traditional reference program (4,820 PD, 6 validation cycles, see model). “Mature” = second or later deployment with a reusable engine and rule libraries.
| Bucket | Baseline | First | Mature | Reasoning |
|---|---|---|---|---|
| Scoping & test design | 800 | 234 | 180 | Usage-based scoping (SCMON/SUSG, ST03N, process mining) removes ~35% of the inventory (assumption; consistent in order of magnitude with the HARMAN/Celonis finding of 34% low-usage custom T-codes). AI-assisted drafting (Cloud ALM Test Management Assistant, Tosca/ECT generation) cuts writing effort ~55%: 800 × 0.65 × 0.45 ≈ 234. Mature: reuse of generated assets → 180. |
| Test data & environments | 600 | 450 | 450 | Better refresh/masking tooling, same 6 cycles: −25%. |
| Test execution | 1,200 | 477 | 400 | 65% of baseline scope; 60% of that automated (Tosca/Worksoft/Opkey-class). Automation build ≈ 1.5 × one manual cycle of the automated part (117). Per cycle: 52 PD manual + 8 PD run/maintenance = 60 × 6 = 360. Mature: existing suite reused → 400. |
| Data validation & reconciliation | 720 | 290 | 290 | DTV / Kyano Validate / Natuvion ARC report diffs: 80 PD setup + 35 PD per cycle for diff analysis × 6. |
| User acceptance testing | 600 | 332 | 332 | Smaller scope (× 0.65) and guided UAT tooling (× 0.85). |
| Defect triage & retest | 600 | 312 | 312 | Fewer cases → fewer defects (× 0.65); AI duplicate/related-defect detection (× 0.8). |
| Documentation, evidence & sign-off | 300 | 180 | 150 | Evidence auto-captured by automation tools: −40% (mature −50%). |
| Total | 4,820 | 2,275 | 2,114 | Factor: first 2.1× (−53%), mature 2.3× (−56%). Fixed part first/mature: 80/80 PD. |
Assumptions
- Usage scoping removes ~35% of the test inventory.
- AI-assisted drafting saves ~55% of writing effort; review stays human.
- 60% of in-scope regression is automatable at reasonable cost.
- Same 6 validation cycles as baseline.
Confidence
Medium-high (±25%). Each lever has published precedents (usage scoping, impact analysis, report-diff validation, AI test generation). The combined factor is my estimate. It stays near 2× because the human oracle (people deciding what “correct” is for each case) remains in every cycle.