Basic concepts of Zuora ProcessAI
To facilitate the smooth adoption of Zuora ProcessAI, the core concepts that you must understand are explained in this article.
Structured data is commonly categorized as quantitative data. It is highly organized and easily understood by machine learning algorithms. You can interpret structured data as the data that fits nicely within fixed fields and columns in relational databases or spreadsheets, and users can input, search, and manipulate structured data relatively quickly using a relational database management system (RDBMS).
Examples of the structured data include account numbers, contact names, bill-to and sold-to addresses, billing dates, credit card numbers, transaction information, and so on.
In general, unstructured data is data that does not have a pre-defined data model or is not organized in a pre-defined manner. It cannot be processed and analyzed using conventional data tools and methods.
In Zuora ProcessAI, unstructured data typically includes:
- Documents (PDFs)
- Log (services logs, IT tickets, Bot logs)
- Conversations such as emails and customer discussion forums.
A process variant is a unique activity sequence from the beginning to the end of the process. You can think of variants as end-to-end use cases.
For example, the following tasks A, B, C, and D are available in your system, where B and C can be used together or interchangeably, while A must be the first task and D must be the last task. Thus, you can get three variants:
- Variant 1: A - B - C- D
- Variant 2: A - B - D
- Variant 3: A - C - D
In reality, your use cases are usually much more complex than this example, so a process map is important as it can provide a bird's-eye view of all processes within your systems. It allows you to effectively identify different variants and understand use patterns, which in turn can help you make decisions and improve your business processes.
A case represents a transaction created or generated for a specific scenario. For example, in the Billing process case, a case represents a generated invoice, and the case ID is the invoice ID. Similarly, in the Payments process, a case represents a created payment, where the case ID is the payment ID. In the Revenue scenario, a case represents a revenue contract, where a case ID is the revenue contract ID.
An activity in ProcessAI is considered an event that is displayed as a node in the process map. For example, if an invoice is posted, this activity corresponds to the “posted” node in the Billing process map.
The Key Performance Indicator (KPI) determines how a process is described and measured. In the context of process maps, KPI describes the measurement for the process between two activity nodes.
Zuora ProcessAI supports the following KPIs:
- Count: The total count of transactions that are processed.
- Time: The mean, median, or a certain percentile of time for a variant or an activity to complete.
The process variants, bottlenecks, and anomalies show the data points with these KPIs.
Anomalies are data points, cases, or activities that were not processed as expected in terms of sequences of activities based on time or amount. They are also called outliers that deviate significantly from the majority of data points. Anomalies might not occur frequently but can indicate potential threats to your business, such as frauds or system errors.
Bottlenecks are the process inefficiencies in the business process. In Zuora ProcessAI, bottlenecks are presented in the “activity → activity” format.
Process correlation is a statistical measurement to describe the extent to which specific activity attributes are related to or influencing each other. Different colors represent different levels of correlations:
|Red correlations||Attributes are highly correlated to time|
|Golden correlations||Attributes are loosely correlated to time|
|Gray correlations||Attributes are not correlated|
Business users can baseline the most common or best practice business process in Zuora ProcessAI. Other process variations can be compared with the baseline for analysis and optimization. There can be multiple versions of the baseline over a period of time and users can use them to see how their process matured or changed during the period.