Process mining is the secret sauce in digital transformation: A recent survey of 1,220 IT decision makers worldwide shows that many organizations have automated the wrong procedures, wasting time and money.
According to the survey, more than one in five respondents (22%) abandoned their automation project completely, while nearly one in three (32%) said the technology didn't work as intended.
Process mining brings together event data from across a set of distributed processes throughout a business to create a "digital twin" that can be analyzed for workflow efficiency. Put more simply, it is advanced AI technology that digs deep into your company's operations to scrutinize procedures and tell you exactly how those can be done better.
While this can be extremely beneficial, traditional process mining does come with challenges that you must understand. It's not a one-size-fits-all technology suitable for every type of process.
Here are three common blind spots that organizations can easily overcome by looking at capabilities that may be underutilized.
1. Schema can't usually capture every useful path
In traditional process mining, most approaches rely heavily on a flow diagram, or schema, of the process being analyzed, such as the one in the figure below. These diagrams are typically augmented with metrics on timing or the count of instances that follow a path from one event to another.
Figure 1. Example of a process flow diagram. Source: ABBY
Schema are useful for understanding what is happening in a distributed process by taking all the data from all systems of record and showing the process flow, how long it takes, and how many instances tend to take one branch versus another.
The limitation with this approach is that, while some business processes tend to follow a small number of path variations, others have much more variability. For many problems of real-world complexity, there are so many routes involved that a flow diagram of all the data would be so complicated that it would be useless. Think of a scribble by a two-year-old child and you've got the picture.
To address complexity and variances, IT providers often offer ways to filter which data is included. This can mean limiting data from various ways a process is completed to follow just the most common path taken. While this can be useful, in a highly variable process, such as a case management situation, filtering out enough data to make the diagram useful risks removing data that is significant to your results, thus invalidating the insights provided. Remember, the most common path taken might be the least efficient.
Process intelligence, the next-generation approach to process mining, deals with this by including the process flow diagram and a wide range of additional numeric analytic tools that provide process metrics and insights, regardless of the variability of the procedures.
The analytics include displays of all specific paths followed, with associated metrics for each, as well as analytic modules to show bottlenecks, costs, timing between steps, event deadlines, various process metrics, and metric histories. You can expect to also see dimensional breakdowns for these.
This use of additional numerical analysis, along with less reliance on the schema diagram, extends the usefulness of process intelligence beyond that of process mining to include all processes, not just those that follow a small number of path variations.
2. Understand and include the manual tasks
The next blind spot involves process steps that are done manually by people on their computers.
When people manually perform a process step on their laptop, they often are logged into an application that will record what is done and when. But many times they are not and instead are looking up data online, changing data in a spreadsheet, or doing some other task that does not create an easily usable trace of the activity.
How people complete their tasks is relevant to how a process is completed. Therefore, many companies are leveraging "task mining" to capture these activities and convert the low-level data on typing/clicking into higher-level events. These events can then be analyzed similarly to the process events captured directly from various systems of record.
The point is not to go into detail on task mining, but simply to emphasize that it needs to be considered as part of the process mining plan. Without task mining, you will be only partially successful at understanding an end-to-end process. There will be gaps, with no data to help understand where people's time is spent or why.
3. Don't forget unstructured data
Another key blind spot with process mining is the access, integration, and understanding of unstructured data. Not all data is available in a log or database generated by a system of record. Capturing manual activities with task mining will provide information on what users are doing on their computers, but that will still leave a lot of data that is in a physical document or in an unstructured text field such as a form.
In some cases, you may not need that unstructured data, but in many others, you will need it for the context it provides to a process. It may provide an analytic dimension to slice and dice your metrics to better understand your process results.
When planning a process mining or process intelligence strategy, take into account what analysis you want to be able to do with your data and if unstructured content is a part of that. You may need to consider intelligent document processing solutions to transform unstructured data. Or you might select a process framework that already has intelligent capabilities to extract information from unstructured content.
Know your blind spots
Process mining can prevent your organization from becoming one of the statistics of a failed digital transformation project—but only if you are aware of your potential blind spots. You want to be able to benefit from behavior and discovery analysis, monitoring, and capability prediction and automatically take action when a process instance requires it.
Your digital transformation projects will also need access to events and data that are often missing in these analyses. These might include manual steps that are not logged in any system of record and unstructured data that is often present as a part of the overall process but is not readily available to be used for context or as an analytic dimension.
Ultimately, the focus needs to be on the insights delivered and on the outputs from process intelligence and how those outputs allow your organization to better achieve your desired results for digital transformation—and not fail.
Keep learning
Get up to speed on digital transformation with TechBeacon's Guide.
Download this free IDC white paper, "Enabling End-to-End Digital Transformation".
See IDC's Futurescape: Top 10 predictions for digital transformation.
How important is digital transformation to your org? Take our survey and find out how you stand next to the competition.
Thinking of making a change? TechBeacon's Careers Topic Center provides expert advice to prepare you your next career move.
