An Intro to Process Mining
Mining ? we will be talking about crypto for sure… Well not really, in a nutshell process mining is the ability to leverage transactional systems data and logs into factful insights for process optimizations.
Before going into the details on how process mining helps getting insights for business operations efficiencies we will start by understanding what a process is and how it can be improved.
A Process is
a series of steps that must be executed within a specific order and/or based on conditions that leads to an objective.
As an example, the process to bake a pizza is composed by the following illustrative steps:
From the process above we can high-light the following:
- the objecting is to deliver a Pizza;
- the standard and main process is composed of 5 mandatory and subsequent steps;
- exceptions to the process can happen — “missing ingredients”
If all activities were perfectly executed and there was no flaws the process would always have the 5 main steps. This is not the case and will never be, there will always be deviations and exceptions to the baseline and those create inefficiencies.
This is where business process management (BPM) tries to optimize and deliver value to organizations by iteratively identify process bottlenecks.
Traditional implementations of BPM typically follow a 5 step approach, mainly driven by the idealization of the perfect business process. Optimizations to the processes are sometimes evaluated without clear evidence of improvements — it is a manual trial and error approach that can take too long until optimizations are found.
This is where Process Mining comes in hand :)
Process Mining work with facts, not perceptions from business SMEs on who the business is running. Process mining extracts meaningful insights on how operations works from transactional systems where all steps are logged.
With these data points, Process Mining is capable of identifying and visually sketch what are the most common process steps and all its different deviations from the main process.
Process Mining methodology follows a 3 step approach:
- Discover — uses transactional data and systems logs to automatically identify the existing business process model and all it variations;
- Conformance— compares the “idealized” business process model with the outcome of the Discovery step to produce a diagnostic that identifies how much the “ideal model” is disconnected from reality;
- Enhance — uses the diagnostic output for the previous steps and produces a new improved refined process model.
In order to instantiate the above process there are 2 main data entry points:
- Transactional Logs — these are data points from databases and logs with real information of what actually is happening on operations.
- Ideal Model — this is the envisioned business process which may not reflect the real process or may not be optimized for efficiency.
Transactional Logs
The data to support a process mining analysis can come from different sources but typically data is obtained from transactional systems and user logs where all steps of the processes are registered. Some examples of these could be data from systems as Salesforce or SAP along with logs from a user computer.
This data can be ingested in real-time or in batch for analysis purposes but regardless on how we access the data the important is to have the mandatory information for analysis. We consider this information as a Event or Transactional Log which must contain at lease the following dimensions:
- Case — this is a unique identifiable number that can be used across all the process steps, example: an Order Number (1001)
- Description — the description of the event, example: Create Order
- Timestamp — this is the date and time in which an event happened
The Event Log table can be extended with more information if available but at least these 3 column are a must.
Ideal Process:
The ideal process is the envisioned process by the Business SMEs, which due to lack of other means for a process analysis they rely on their own view to define the processes:
“I can tell you, this is exactly how we operate”
“we are as efficient as possible”
“it is impossible to improve this further”
“there is no better way to do it”
The purpose of Process Mining is a mean to detect the above statement and provide clarifications on:
what’s actually happening ?
how to improve it further ?
Process Mining End to End Process:
All of this is iterated over and over again to continuously deliver improvements by analyzing and recommending optimizations.
[To Be Continued — with pratical implementation]
