The challenges of digital transformation
Thousands of large companies around the world have begun the process of digital transformation, but only a few succeed in carrying it out. Large budgets are laid for this, but, unfortunately, it is not always possible to achieve the desired result due to the lack of a strategy.
To achieve a successful transformation of business processes, it is necessary to set two big goals: increasing the efficiency of business processes and reducing risks. This requires:
- Detect the problem
- Implement changes
- Track Solution
- React to deviations
At the moment, a popular solution for implementing changes is the RPA tool, which allows you to automate routine user actions. As for the other three points, this will require process mining or intellectual analysis of business processes.
Process Mining focuses on the discovery, analysis and optimization of business processes based on data from event logs, presenting the missing link between classic business process analysis using their models and data mining.
The process mining tool allows you to capture real-world data from popular ERP, CRM systems and databases of end-to-end business processes in procurement, finance, claims management, contact centers, etc. (SAP, Oracle, Salesforce, ServiceNow), visualize them to detect bottlenecks, resource inefficiencies and exceptions; and, finally, monitor changes in the process after its optimization, including through automation.
You might say, "The processes in my company are well described and the employees follow the instructions clearly." And ideally this should be so, but in practice we can observe the opposite picture. I will give a simple example, by the way, it is well described in the article on Habré "Street dirt and the simulation of pedestrian traffic": your yard probably has paths made of tiles or asphalt that were designed by the local government and convenient paths trodden by your neighbors that allow you to get to the desired place along the shortest path.
Photo from the resource: https://habr.com/ru/post/257563/
A similar situation occurs in the company's business processes. People tend to go not according to a given algorithm if they feel its inefficiency and “trample” changes in the form of additional systems, unnecessary actions and requests.
Finishing my analogy, I would like to note that process mining allows you to centrally detect these very "paths" in business processes and make decisions about their change or automation based on facts, not intuition.
Process mining and automation cases
Auditing and improving existing metrics
A large retirement, insurance and management company in the Netherlands needed a deeper understanding of their contact center automation in order to understand its impact on their business processes. The company deployed UiPath Process Mining, conducted end-to-end contact center analytics, and identified issues that resulted in additional work and long customer waiting times.
Based on insights from UiPath Process Mining, the company has adjusted the automation of the contact center. This increased the quality and speed of communication with clients, and also greatly reduced the waiting time for the operator. In general, it turned out that 80% of all work could be standardized, and due to this, the costs of contact centers were reduced by 568 thousand euros.
Operational decision support
One of the world's largest automotive suppliers was looking for ways to improve internal processes. After consulting with UiPath, the company decided to use UiPath Process Mining to streamline its corporate procurement process to identify control risks, supplier duplication and factory delays in regions around the world.
After implementing UiPath Process Mining, the manufacturer managed to achieve complete transparency of procurement: gain a new understanding of the root causes of delays, inefficiencies and potential audit risks. He found new ways to improve procurement results at multiple factories and identified the true costs of certain management decisions. This prompted top management to think about and evaluate the possible effects of the reorganization and optimization of production.
Data mining and deep understanding of processes
Using data mining, you can extract useful information about goods and consumers: segment consumers. But if you understand that each transaction is the result of the process of interaction between the consumer and the company, and by studying these processes themselves, you can get a deeper understanding of what is happening: information about how customers appear, how they make decisions about working with the company, what affects the it is the decision, and ultimately why customers leave.
A large telecom company from the Netherlands needed to improve the transparency of its corporate purchases. With 6.3 million fixed-line subscribers, more than 33 million mobile subscribers and more than 2 million internet customers in five countries, they wanted to find a solution to control risks and identify ways to reduce costs.
The operator deployed UiPath Process Mining to gain insight
and eliminate intuitive process evaluations as well as manual data transfers. He used UiPath Process Mining to evaluate over 200,000 different items, including purchase orders and invoices.
The result was a 20% reduction in labor costs, a 29% reduction in invoice processing time, and improved cost predictability and improved supplier relationships.
Customer experience
Process mining is an effective and powerful tool that can be used to analyze a company's business processes and then transform its experience to remove bottlenecks and minimize costs. However, for real transformation, it is not enough just to understand the details of all processes; a number of conditions must still be met.
Ekaterina Sabelnikova, Philips Innovation Consultant, talks on her LinkedIn blog about the main lessons learned from process mining in the company.
Focus on the essentials
There are a lot of optimization processes in medium and large organizations, so it is important to choose strategically significant ones that will lead to the desired goals. The simultaneous improvement of processes on all fronts will require excessive consumption of resources, and the quality of such work will naturally decrease. Focus on a couple of the most promising key areas and track their performance.
Identify Stakeholders
Only those employees who are really interested in the solution and want to get to the bottom of the problem are really willing to spend additional efforts on improving bottlenecks in processes. These will be those people whose work is directly affected by the non-optimal implementation of a particular process.
Philips takes a top-down approach, starting by learning about the process without clearly stating a problem or business problem. Once the problems in the process are identified, the group in charge begins to communicate with the staff who are carrying out the process. It happens that people who work on this process every day do not themselves feel discomfort from it and are not inclined to change. Therefore, you need to find employees who really feel this "pain" and can influence these colleagues.
Securing senior management support and approval
First, some improvements require structural transformation within the organization. In this case, senior management can help shape the ad hoc project and allocate resources for its implementation.
Second, senior management can make strategic decisions that affect the company's operations. Process improvements often go hand in hand with trade-offs.
Finally, management can position process mining as a tool to help meet KPI targets. Promote the involvement of all levels of the organization in change and the use of process mining as a daily management tool.
It is important to understand that process mining is not a “silver bullet” that will solve all your problems, optimize processes and make your operations efficient and productive. Remember, identifying problems and potential areas for improvement is just the beginning of a digital transformation in a company .
As with any system, process mining has a number of limitations:
- Adequacy of displaying the progress of a real business process with data from the information system logs;
- The need to interpret the analysis results.
We will face these limitations in practice, when further, using the example of UiPath Process Mining, we will analyze the entire process of introducing business process analytics into an organization.
UiPath process mining helps make evidence-based recommendations to improve critical processes
UiPath Process Mining Deployment Guide
Forming an event log
To create an event log, you need to define data sources for it, as a rule there are no more than 2-3 of them. Even if a company has SAP in UiPath and its own ETL, in the real world this will not be taken advantage of by the company's integration strategy and the presence of QCD / KShD / DataLake or other solutions that control data flows.
We obtain the initial data from the data warehouse or database replicas and the ETL tools adopted in the company, then collect / simplify them and place them in the staging area for UiPath. UIPath can take data from any relational database that has an ODBC driver. Then, from the staging tables, we collect the event log, with which the UiPath algorithms themselves will work.
After the event log is created, it is necessary to work out the logic of its structure and collect all the data in one place.
Assembling the event log
Technically, assembling the event log is just select / insert from a bunch of tables into one or two (two, when there are attributes that do not change from step to step of the process). The non-obvious part related to the design of the structure of such a log: how to transform the change in the status of an object or the fact of a change in the value of an attribute into a specific step. This is the task of a business analyst, there is no universal methodology, there is a vendor template and an integrator's know-how.
It is important to understand that an incorrectly selected composition of steps and a level of process detailing will not allow solving the problems of analysis or decision support. And yet, there is no 100% digitalization of the process anywhere. It means that you will not reflect some of the steps that are in the process in the journal. If you remember that you are solving the problem of increasing indicators, and not getting a beautiful picture, this is not scary. And if you do not remember, then you can decide that process mining is a useless implementation.
After you have worked out the structure of the log, you need to configure the process mining UiPath itself, where you need to correlate the tables and their fields with the attributes of the event and the chain of events.
Configuring Process Mining UiPath
To do this, you need to fill in the configuration file, usually there is all the documentation for this. Since everyone will have their own steps in the process, it makes no sense to write an example.
It is important to note that UiPath has a very useful anonymization feature. Which, for example, changes the counterparty “Best company” in all places to “Counterparty 1.” This can, of course, be done by creating a temporary reference table or by writing a very tricky query, but this feature saves a lot of time. In general, depersonalization of results is a very important thing If you analyze the processes responsible for financial flows. It may turn out that business analysts cannot see trade secrets, but for analysis they are important. This is where anonymization comes in.
After all the data is in its place, you need to solve a technical problem - write the KPI calculation formulas.
We spend reporting analytics
Analyze reports technically easy, but to interpret their results is not easy - it is the sphere of responsibility of the analyst. The drag-and-drop UiPath method greatly simplifies its work: pulled out a widget, set it up and look at the data. What exactly and how to analyze depends on the goals of the analysis. From a technical point of view, it is important that the visualization for the analyst is prepared by the report designer, and then his work begins.
We create dashboardsBased
on what the analyst finds, it is necessary to design dashboards for regular work according to analytics requirements, write access rights and perform a whole series of routine work, including styling according to the customer's standards.
After the initial analysis has been carried out, the requirements for operational control can be formulated.
Setting up operational control
In UiPath, it is possible to write a control rule (for example, more than 5 edits of the delivery date when agreeing on dates is bad) as a function that is used when loading a fresh portion of information. UiPath can do incremental data loading, which means that ETL from past caps can be written so that it can only load new data.
The result of the function can be used in dashboards to mark suspicious cases or send it to the UiPath Action Center to record the procedure for its elimination. By the way, the integration with ML also works - you can call programs in R or Python, passing data to them as input and recording the results of the work.
As you can see, the work here is always iterative. This is bad when planning and estimating the scope of work, because it is impossible to accurately estimate whether another iteration will be required or not. But this is not a property of UiPath, it is a property of all such products. This is good because it becomes possible to do what is really needed and what will be used, and not what is fixed in the requirements before the start of work.
There is a technical problem here - how to organize layout control. From the very beginning, everything is correct with UiPath: the configuration of settings and dashboards lives in git. You can do it locally, you can do it in a corporate one, you can even go to gituhub. In other solutions, somewhere there is no versioning at all (for example, the configuration of the data model), but somewhere it is achieved in the good old way “add the date or version number to the file name”. Technically, a lot of headaches go away with verstan control and the sprint duration is reduced.
UiPath product solutions make it easy and comfortable to implement process mining in a company. The criterion for the success of the project will be the methodology that will allow you to go from defining the goals and boundaries of the analysis to identifying the reasons for negative trends in the characteristics of the process and creating a tool to support operational decisions.
Any end user with basic developer skills can use the UiPath platform tools. The second criterion for the success of process mining projects is the strong expertise of process analysts. The maximum effect from optimization of processes is obtained not from the formation of recommendations, but from the systematic adherence to them, and UiPath tools will help here with the technical implementation of process mining in the company.
The Everest Group annually evaluates process mining technology vendors based on their market impact, vision and product capabilities. In the latest research from 2020, UiPath was recognized as the leader in process mining among other major vendors.
conclusions
The necessity and importance of digital transformation is now recognized by many companies, but not everyone understands how to implement it correctly, and not just waste budgets. The basis for the digital transformation of an organization is the combination of robotization and automation (RPA) solutions with tools for deep analytics of business processes (process mining). The latter allow you to objectively assess how the processes in the company actually take place, identify bottlenecks and determine the points for future transformation.
The implementation of process mining requires a competent approach, and successful implementation depends on various factors, including the management's vision of transformations and the effectiveness of internal communications in the company. In addition to the process mining tools themselves and specialists who can configure them, a systematic approach is also needed to implement the recommendations received by analysts. Simply put, the process mining tools themselves are only a third of the success, strong expertise of process analysts is the second important component, and, finally, competent management decisions and work with the team are another necessary component.