10 Principles of Attention Allocation

10 Principles of Attention Allocation

Attending to the things that matter is the foremost task of any manager. This means deciding which things matter and how much time to attend to those things. And as we all know the time to decide is shrinking dramatically while the number of to be taken decisions is growing by magnitudes. We at Trufa are set out to help business managers with these decisions. We are trying to mimic human reasoning with machine learning. For this we are applying following principles in our algorithms: Attention allocation algorithms must be comprehensible by human beings. Algorithms people don’t trust in will not be accepted. Attention allocation algorithms must produce dependable results. Algorithms never produce 100% precise results. But their margin of error must be negligible. Attention allocation algorithms must produce reproducible results. Algorithms must be reliable. Attention allocation algorithms must be robust against outliers. Human beings tend to overly turn to outliers though often these outliers are unmanageable. Attention allocation algorithms must be unbiased. Among others multi-variate regressions make assumptions about the applicable variates. Attention allocation algorithms must be diligently applied. In the 80ties expert systems were the cure for everything. Nowadays machine learning is overrated. Attention allocation algorithms must work in real-time. Decisions are to be taken ever faster these days. Attention allocation algorithms must work on very large data sets. ERP systems are growing day by day. And the digital transformation is accelerating this. Attention allocation algorithms must work for various businesses in various industries. Trustworthiness beats individualism. Attention allocation algorithms must work across multiple ERP systems. Non-matching document ids and master data must not throw of the...
Process Mining on Shaky Ground – or – The Fallacy of (SAP) ERP Event Logs

Process Mining on Shaky Ground – or – The Fallacy of (SAP) ERP Event Logs

Process Mining is highly popular. Because customers want to learn about their business. In detail. Unfortunately, this high level of customer interest has lured vendors into taking technical shortcuts which are questionable at least. The typical approach in process mining is to look out for so-called event logs. “Event logs: To be able to apply process mining techniques it is essential to extract event logs from data sources (e.g., databases, transaction logs, audit trails, etc.).” (http://www.processmining.org/logs/start) In case of technical systems such as database or transaction management systems event logs are the basis for ensuring their transactional integrity. “Examples from double-entry accounting systems often illustrate the concept of transactions. In double-entry accounting every debit requires the recording of an associated credit. If one writes a check for $100 to buy groceries, a transactional double-entry accounting system must record the following two entries to cover the single transaction: 1. Debit $100 to Groceries Expense Account 2. Credit $100 to Checking Account A transactional system would make both entries pass or both entries would fail. By treating the recording of multiple entries as an atomic transactional unit of work the system maintains the integrity of the data recorded. In other words, nobody ends up with a situation in which a debit is recorded but no associated credit is recorded, or vice versa.” (https://en.wikipedia.org/wiki/Database_transaction) The transactional integrity can only be preserved if and only if the respective event log is 100% accurate and complete. Hence event logs in transactional systems are highly reliable for subsequent process mining as for example performed by Splunk. In case of business systems like SAP ERP such...