Training Objectives

What to expect from this day

Agenda

Outline of the day

1. Introduction to Process Mining & OCPM
2. Data Integration — advanced topics (Sharing, Delta, Replication Cockpit)
3. Objects & Events — modeling, Foreign Keys, Overwrites, Versions, and Perspectives
4. Troubleshooting — common errors in Objects, Transformations, Perspectives, and Views
5. PQL & Views — differences from SQL, essential commands, and performance cheats
6. Action Flows — orchestration and integrations (specific reinforcement)
7. Process Orchestration — Orchestration Engine, Forms, Events & end-to-end automation
8. Wrap-up & Q&A

From Case-Centric to Object-Centric

The conceptual shift in one picture

What changes in practice

When Celonis migrates from Case-Centric to OCPM

The 4 layers of Celonis Platform

From raw data at the bottom, to user interaction at the top

Studio vocabulary — Cheat-sheet

The names you'll hear all day long

Where AI & Agents fit in

Three levels of intelligence inside the platform

Data Integration Overview

From SAP tables to Celonis objects — the pipeline journey

Extraction Strategies

Full, Delta, and Replication — when to use each

Pipeline Health & Governance

Replication Cockpit + Sharing — keeping the pipeline safe and collaborative

Replication Cockpit — Reading the signals

What each column tells you about pipeline health

Transformations Deep Dive

Adding intelligence to raw data

Hands-On — Expose SAP Flag as Object Attribute

Adding EBAKZ (Purchase Requisition Closed) visibility for PR analytics

SAP ECC → S/4HANA Migration

What changes from a Data Integration perspective — future-proofing

• Tables that change name or structure — e.g., BSEG/BKPF coexist with ACDOCA; MSEG with MATDOC.
• Extractor: the S/4HANA connector already provides its own templates — pay attention to the version used today by MPWA.
• Impact on Transformations: SELECTs need to be reviewed; old fields may become compatibility views.
• Recommended strategy: validate end-to-end (Extractor → Data Pool → Transformations → Knowledge Model → View).

Objects and Events — Introduction and Context

What they are, how they relate

Relationships: Foreign Keys vs. Dedicated Tables

Modeling 1:N and M:N correctly

Event Ordering & Determinism

Why timestamp alone is not enough

Cardinality & Relationship Design

When to use each pattern

1:N — Foreign Key

1 PO Header
  ├─ PO Item 1 (FK: EBELN=5001)
  ├─ PO Item 2 (FK: EBELN=5001)
  └─ PO Item 3 (FK: EBELN=5001)

M:N — Junction Table

PO Item ─┐
          ├─ Relationship (junction)
Invoice Item ┘
Avoids duplicates on BOTH sides

❌ Anti-pattern

Overwrites & Custom Attributes

Customizing without breaking the template

Perspectives Under the Hood

How a perspective materializes the "lens"

What a Perspective Does:

🔍 Validations During Materialization:

Versions, Deploys & Perspectives

The lifecycle of Knowledge Model changes

🔄 Typical Workflow:

Augmented Attributes — Deep Dive

Extending objects with data that doesn't live in SAP

💡 Augmented vs. Native

🎯 MPWA Examples

Hands-On — Augmenting the PR Item object

Adding triage attributes so Buyers can record decisions in the View

Troubleshooting — Objects & Events

Symptoms, root cause, and solution

Troubleshooting — Transformations

When the pipeline complains

-- Before (problematic)
SELECT a.id, b.name
FROM ekko a LEFT JOIN lfa1 b ON a.lifnr = b.lifnr;

-- After (defensive)
SELECT
  COALESCE(a.id, 'UNKNOWN') AS id,
  b.name
FROM ekko a LEFT JOIN lfa1 b ON a.lifnr = b.lifnr
WHERE a.id IS NOT NULL;

Troubleshooting — Perspectives

The 3 most common errors

Troubleshooting — Views

Most frequent errors in MPWA Views

Hands-On — Troubleshooting a View

Read the errors, find the root cause, fix the Free-Text Requisition Processing App

SQL Window Functions vs. PQL

How to replicate SQL window logic in Celonis

SQL Date Functions vs. PQL

Date manipulation — native support in Celonis

PQL — Basic commands

The ones that appear in almost every view

MPWA Real Case — INDEX_ORDER

Filtering directly in Get Rows 🔗

Applied solution: use INDEX_ORDER to generate a position for each row within each category, and filter directly in the Get Rows of the Action Flow.

-- Position of the case ordered by date, within the status
INDEX_ORDER(
  "PO_Header"."Document_Date",
  "PO_Header"."Approval_Status"
)

-- In Get Rows, filter relevant records
FILTER INDEX_ORDER(...) <= 50;

MPWA Real Case — RUNNING_SUM & RUNNING_TOTAL

Accumulating over time 🔗

-- Cumulative Net Order Value by document date
RUNNING_SUM(
  "PO_Header"."Net_Order_Value",
  ORDER BY "PO_Header"."Document_Date" ASC
)

-- RUNNING_TOTAL — equivalent, with partition
RUNNING_TOTAL(
  "PO_Header"."Net_Order_Value",
  ORDER BY "PO_Header"."Document_Date" ASC,
  PARTITION BY "PO_Header"."Vendor"
)

MPWA Real Case — PU Functions + BIND

AFP View — displaying the last Invoice number 🔗

-- Last Invoice Number per PO Item, via m:n relationship
PU_LAST(
  "o_celonis_PurchaseOrderItem",
  BIND(
    "o_celonis_RelationshipThreeWayMatch",
    "o_celonis_IncomingMaterialDocumentItem"."SystemIncomingMaterialDocumentNumber"
  )
)

Why BIND? The objects PurchaseOrderItem and IncomingMaterialDocumentItem are not directly related on AccountsPayableExtension perspective, but they are indirectly related by RelationshipThreeWayMatch object (via a SQL perspective, this table is the junction table between both previous Celonis objects).

MPWA Real Case — PU_STRING_AGG_DISTINCT

Condensing data at the PO Header level

PU_STRING_AGG_DISTINCT(
  "PO_Header",
  "PO_Item"."Broad_Category",
  ', '
)

MPWA Real Case — LOOKUP

PO Terms Checker App — justification corrections 🔗

PU_LAST(
  "o_celonis_PurchaseOrder",
  LOOKUP(
    "O_CELONIS_PURCHASEORDER_AUG_BUYER_JUSTIFICATION",
    "o_custom_CelonisMembers"."NAME",
    (
      "O_CELONIS_PURCHASEORDER_AUG_BUYER_JUSTIFICATION"."CHANGED_BY",
      "o_custom_CelonisMembers"."ID"
    )
  )
)

MPWA Real Case — INTERPOLATE

Users & Value Tracking — continuous line even when granularity changes 🔗

INTERPOLATE(
  COUNT("User_Activity"."Activity_ID"),
  "User_Activity"."Activity_Date",
  '${time_range}'           
)

Performance Cheats — PQL & Views

Top tips for fast views

1. Pre-aggregate in the Transformation whenever possible — don't leave the heavy lifting for the view.
2. Avoid redundant FILTERS — one well-placed filter is worth more than several scattered ones.
3. Use PARTITION BY in RUNNING_* and window functions to reduce the calculation scope.
4. Be careful with large DOMAIN_TABLES — they explode the Cartesian product.
5. Limit columns in analytical tables — the front-end renders everything.
6. Cache KPIs: define KPIs in the Knowledge Model and reuse them, don't recalculate.

Hands-On — Build & Extend the Free-Text PR View

Add KPIs, an analytical table, and a detail panel for Material recommendations

Hands-On — Historical context with PU_STRING_AGG_DISTINCT

Show Buyers where each recommended Material has been used recently

Action Flows — Recapping

What to highlight from the Process Orchestration training

Hands-On — Daily digest Action Flow for Free-Text PRs

Email Buyers their Top 20 oldest PRs every morning · using INDEX_ORDER

Process Orchestration — What it is & why it matters

Process Orchestration — Core components

Hands-On — Free-Text PR "Need Info" orchestration

Close the loop: Buyer asks → Requester answers → Process resumes

Day's Recap

Our key takeaways

Q&A — Open Space

Which MPWA scenarios do you want to break down?

• Any recurring error in transformations that hasn't been solved yet?
• Any view with performance issues today?
• Use cases for Action Flows that are in the queue?
• Doubts about the post-migration roadmap for SAP S/4HANA?
• Augmented attributes — other possible uses?