How SeaLights for ABAP Works

SeaLights for ABAP operates in two layers that activate in sequence.

Day Zero

The first, Day Zero, delivers change impact analysis and test recommendations from structural analysis alone, before a single test runs.

SeaLights maps SAP executions to their associated code objects, and links those objects to the tests in your test repository. When a transport is applied, SeaLights identifies which objects changed and recommends the tests mapped to those objects. This model works from day one, with no execution history required, but because it relies on structural mapping rather than observed behavior, it may produce a broader recommendation list than strictly necessary.

The secon, Day One Onward, learns from actual test execution to verify coverage and sharpen every future recommendation.

---

The Two-Layer Architecture

Transport imported into SAP system
            │
            ▼
┌───────────────────────────────────────┐
│             DAY ZERO                  │
│                                       │
│  Step 1 — Build the Dependency Map    │
│  Step 2 — Rank Most-At-Risk           │
│  Step 3 — Build Optimal Test Set      │
│                                       │
│  Output: Change Analysis,             │
│          Tests to Run, Test Gaps      │
└───────────────────────────────────────┘
            │
            ▼
       Tests Execute
            │
            ▼
┌───────────────────────────────────────┐
│         DAY ONE ONWARD                │
│                                       │
│  Observe what was actually hit        │
│  Confirm or drop predicted links      │
│  Refine the model for future cycles   │
│                                       │
│  Output: Verified coverage,           │
│          Smarter future recommendations│
└───────────────────────────────────────┘

---

Semantic Model

Statistical Model

As test executions are reported back to SeaLights, the system learns which tests actually exercise which code paths. Over time, this observed data refines the semantic mapping, improving precision and reducing the recommendation list to only the tests most likely to catch a regression. The statistical model builds on top of the semantic model - it does not replace it.

Every integration level starts with the semantic model. The statistical model develops progressively as execution data is reported to SeaLights, improving recommendation effectiveness over time.