What is an order-to-cash process?
Order-to-cash (O2C) is the end-to-end process from order entry to payment clearing. It covers order capture, configuration and release, delivery and invoicing, payment and dunning, and complaints and rework. In the operational data O2C is not a linear flow but a network of business objects such as customer order, item, configuration, delivery, invoice, payment and complaint.
What does touchless mean in order-to-cash?
An order counts as touchless when it runs through the process without manual intervention. The touchless rate measures the share of such orders – 46.37% in the synthetic demo dataset, i.e. 1,978 of 4,266 consistently assessable customer orders.
Should the touchless rate be as high as possible?
Not at any price. Some manual steps only cost time, others prevent costly errors. In the O2C example, the complaint rate of unreviewed technical orders rises markedly with the number of special values. What matters is therefore which orders need a human and which do not: as much touchless as possible, as much effective review as necessary.
What does the OTIF rate measure?
OTIF stands for “on time in full”: an order was delivered on schedule and complete. In the demo dataset, 3,708 of 4,266 consistently assessable orders meet this criterion – an OTIF rate of 86.92%.
Why does Noreja treat deviations as error patterns rather than variants?
In event-log-based approaches every deviation quickly appears as a variant of its own; the number of variants grows and the analysis becomes confusing. Noreja describes the expected flow in the causal perspective and classifies violations – such as skipping, wrong order or rework – as error patterns. In the O2C example, 197 of 340 technical orders changed before release go straight to release without the intended review.
Is an error pattern already the cause of a problem?
No. An error pattern shows where the expected flow is violated. It is the starting point of root cause analysis, not its result. In the O2C example the missing review is first a relevant path of analysis – only the cross-check shows that it is above all the risk profile of the special configuration that separates good from bad cases.
Why is a change before release not automatically the cause of complaints?
Because the cross-check does not generally confirm the blanket hypothesis. For delivered technical orders without review, the complaint rate rises above all with the number of special values – from 2.23% without a special value to 21.51% with one and 35.82% with two or more. Within the same number of special values, the change shows no consistent picture.
What role does the AI assistant Minerva play?
Minerva answers business questions in dialogue, structures complaint reasons, forms comparison groups and separates documented findings from cause hypotheses. It connects process data, company context and causal process mining. The decision about measures stays with the process owner and the business.
Why does AI need causal process mining?
Generative AI can summarise patterns quickly but does not by itself distinguish correlation from cause. AI-supported process decisions need the right business granularity, relationships between business objects, company context and an analysis that tests cause hypotheses. Automating a process on the basis of the wrong analysis only makes it wrong faster.
How does a finding become a measure?
In the Business Impact Board, findings are rated and prioritised by expected impact, implementation effort, risk and uncertainty. The chosen measure is implemented as a pilot; afterwards touchless rate, complaint rate, OTIF and cost are measured again with the same process data.
Are the figures from a real company?
No. All figures come from a synthetic order-to-cash dataset. They illustrate the analysis method and are not production figures of any customer.
