Indu Learning Platform
Course/Unit 7 · Capstone and career
Lesson 7-3

Interview patterns for technical MDG

Four question types, one answer structure (clarify, draw, name, risks, result), worked example answers, and STAR stories built from your real projects.

Watch firstUse the STAR METHOD for success in your JOB INTERVIEW · English with Ronnie · EnglishLessons4U with engVidOpen on YouTube ↗

The four question types

Scenario tests end-to-end system behaviour. Design tests trade-offs. Debug tests methodical investigation. AI tests using new tools without breaking governance. Prepare two answers per type: one from real projects, one from the course. Behavioural questions get STAR stories.

Technical MDG interviewScenarioCR stuck inworkflowApproved but notreplicatedDesignSupplier onboardingWhere to put avalidationDebugDuplicates inreceiverIDoc errors aftergo-liveAIWhere to use an LLMHow to avoid wrongmerges
Two prepared stories per branch covers most first and second rounds.
Try it · Sort

Sort these interview questions by type.

Scenario
Design
Debug
AI
Click an item, then a bucket (or drag it).

One structure for every technical answer

Same five moves every time. It keeps you calm, shows seniority and gives natural points for follow-ups. On video, draw boxes on paper and hold it up.

Step through · 1 of 5
ClarifyDrawNameRisksResult
Ask one or two scoping questions: hub or co-deployed, which entity, volumes, which receivers. It shows the answer depends on context.
Clarify, draw, name, risks, result.
Try it · Put in order

Put the five moves of a technical answer in order.

  1. 1Clarify: ask about landscape, entity and volumes
  2. 2Risks: what could go wrong and how you'd check
  3. 3Name: CR type, BAdI, DRF model, transactions
  4. 4Result: close with an outcome or metric
  5. 5Draw: sketch the flow as boxes and arrows

Worked example: scenario

'A supplier CR was approved but never reached ECC.' Clarify: hub, IDoc, one CR? Then walk the path: is the CR finally activated? Did DRF select it (DRFLOG)? Replicate manually (DRFOUT), follow the IDoc (WE02), read SLG1. Risks: DRF filter, missing key or value mapping, manual replication mode; never resend blindly in production. Result: a checklist like this kept your issue rate below 5%.

CR activated?CR statusDRF selected?DRFLOGManual testDRFOUTMessage statusWE02 / SRT_MONIApp logSLG1Fix filter / mappingDRFIMG
Walk down the replication path until you find where it stops.

Worked example: design

'Design supplier onboarding that prevents duplicates.' Clarify volumes, requesters and receivers. Name: a dedicated CR type with rule-based workflow, the MDG duplicate check at create, BRFplus for simple field rules, a BAdI for harder checks, DRF with key and value mapping. Risks: an over-strict check annoys requesters; small suppliers may lack tax numbers. Result: duplicate rate and CR cycle time before and after.

RequestFiori app / portalDuplicate checkmatch profile at createValidationBRFplus + BAdIApprovalrule-based workflowActivation + DRFkey / value mapping
Stop duplicates at the door, then govern and replicate.

Worked example: debug

'The receiver has duplicate business partners after go-live.' Clarify: created by replication, or migrated? Test hypotheses in order of likelihood, each with evidence. Fix through MDG, never directly in the receiver, or the next replication brings them back. Result: a root-cause note plus prevention (duplicate check at create, or consolidation for the backlog).

HypothesisCheckWhereDuplicates already in MDGGroup BPs by tax numberSQL on DFKKBPTAXNUMMigration loaded the same suppliertwiceCompare source IDs per BPMigration logs, mapping fileKey mapping missing, soreplication sent a createLook up mapping for BPKey mapping toolsSame BP replicated twiceRead replication log for BPDRFLOG, WE02Receiver created it locallyCheck creator in receiverReceiver change log
Most likely first; each row has evidence you can show.
abap
" Quick check: tax numbers used by more than one Business Partner
SELECT taxtype, taxnum, COUNT(*) AS bp_count
  FROM dfkkbptaxnum
  GROUP BY taxtype, taxnum
  HAVING COUNT(*) > 1
  INTO TABLE @DATA(lt_dupes).

cl_demo_output=>display( lt_dupes ).

Run in a sandbox or with read access. It answers hypothesis 1 in seconds.

Worked example: AI

'Where would you use AI in MDG, and where not?' Clarify: SAP's own features or side-by-side services, and data restrictions. Risks: hallucinated reasons, data leaving the landscape, silent prompt changes. Controls: validated structured output, logged model and prompt version, data minimisation, acceptance-rate tracking. Result: point to the capstone.

Use AI toExplain duplicate pairsSuggest rules from profilingSummarise replication errorsDraft specs and test dataDo not let AIApprove change requestsMerge or delete recordsWrite to the active areaDecide regulated data alonevs
AI suggests on the side; governance stays on the main path.
Try it · What would you do?

The interviewer says: 'Design supplier onboarding that prevents duplicates.' How do you open?

STAR stories from your real projects

Situation, Task, Action, Result. About two minutes each, most time on Action, end with a number. Prepare four or five and map each to the questions it answers. Share only industry and scale, not confidential client details.

Your STAR storyQuestions it answersTwo go-lives with 15 sourcesystemsPost-replication issues keptbelow 5%Custom LSMW programs for atricky objectRCA on MM enterprise structure(Cognizant)UFT regression suite(Cognizant)A complex delivery underpressureHow you improve qualityA technical problem you solvedA time you found a root causeHow you made work repeatablealso works
Five stories cover most behavioural questions.

Remember

  • Most technical MDG questions are scenario, design, debug or AI.
  • Clarify, draw, name, risks, result: the same five moves every time.
  • Prepare five STAR stories from real projects, each ending in a number.

Check yourself

What is your first move on a design question?
A CR is approved but the receiver has no change. What do you check first?
Which is an appropriate use of an LLM in MDG?
In a STAR answer, where should most of your time go?