Reproducible reference run
What one run actually produces
Most claims about tools like this cannot be checked. This one can. Below is a complete run over a legacy ABAP program that ships in our repository — every finding, and the split that tells you how much of the work the tool takes off your desk.
Download the file, run it yourself, and you should see the same numbers.
The analysis itself takes milliseconds. That is not the point, and we do not claim it saves you days — what takes time is the decisions, and those stay with you. The point is the split below: it tells you which decisions you still have to make.
Where the work lands
the tool settles
Points at a released SAP successor — SAP’s own where it names one, the curated mapping otherwise. The finding says which. You review it; you do not have to find it.
your decision
Transformable, but somebody has to weigh business intent against the target design. The tool shows the evidence and stops.
stays hand work
Out of reach for any generator. Flagged and isolated rather than guessed at, so nothing false reaches your draft.
In this file: dynpro, native-sql.
By severity
The engine also recommends a target route for this program: Side-by-Side (SAP BTP). That recommendation is derived from the findings, not from the AI layer — you can see the reasoning in the product.
What this run is not
- One file, not a codebase. It is a single reference program we wrote, deliberately dense with legacy patterns. Your ratio will differ — read the file and judge for yourself how close it is to what you have.
- Synthetic, not a customer system. We publish it precisely because we can: no customer code is involved.
- Not a promise about your result. It is a demonstration of the method and of where the boundary sits — nothing more.
Check it yourself
The exact file this run used. Load it into the free analysis and compare — that is the whole reason it is published.