Validate Your Data

This page applies to business users. It describes how to validate the data in your process, so you can verify and potentially refine the data preparation in collaboration with data stewards and developers.

Overview

The quality of the insights you see in Process HQ relies on the quality of the data in your process. When you add a process to Process HQ, your next step should always be to validate that the data matches your business knowledge of the process.

This step lets you know that your organization has adequately prepared the process data for analysis, so the insights Process HQ shows you will be accurate and relevant. Identifying and eliminating discrepancies early keeps you from spending time on irrelevant or incorrect data as you and your colleagues explore the process.

To validate your data, we recommend you perform the checks described below. If you identify any inconsistencies or inaccuracies in your data, you should work with your data steward and developer to troubleshoot process data collection, process preparation, and process filters.

Check cases

Check the following details to validate which cases are included in your process.

Case count

Check the case count in the header of the Process Insights page.

For example:

In this example, you might have expected a total of 3,000 cases. A data steward may need to filter out cases for a specific region or time period, or a developer may need to determine if duplicate cases are being tracked.

Date range

Check to make sure the process data covers the expected date range:

1. Go to the Time Period setting in the header of the Process Insights page.

   For example:

   

   Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

2. Select Custom Date Range.
3. Click the Any - Any dropdown list.
4. Set the expected start and end dates. Process HQ applies the date filters.
5. Review the case count in the header, and determine if it is higher or lower than expected.

Specific cases

If you are expecting specific cases to appear in the process, you can verify that they are present by going to the Case List tab on the Process Insights page and entering a Case ID.

Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

The case ID uniquely identifies the specific occurrence of the business process. It matches the primary key for a specific record in the case record type.

Check KPIs

Go to the PINNED VIEWS section of the Process Insights page, and check the following KPIs to validate which cases are included in your process.

Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

Average case duration

Go to the PINNED VIEWS section of the Process Insights page.

Is the average case duration KPI higher or lower than you might expect?

For example, based on your business knowledge, you might expect a Financial Customer Onboarding process to take about a week to complete. If you see that the average duration is 1 day, you might suspect that data is missing from the process, which a data steward or developer can help you to track down.

Median case duration

Go to the PINNED VIEWS section of the Process Insights page.

Is the median case duration KPI significantly different than the average case duration KPI?

Cases with significantly shorter or longer duration than other cases in the same process represent exceptions to how your process operates most of the time. The median case duration KPI accounts for these exceptions by showing the average duration of the majority of cases, rather than the average duration of all cases.

If the median duration KPI and average duration KPI are dramatically different from each other, you may want to identify the exceptions and work with your data steward and developer to filter those cases from the process altogether.

Automation rate

Go to the PINNED VIEWS section of the Process Insights page.

Is the automation rate KPI lower or higher than expected?

For example, you might be aware that multiple activities in your process are automated, but you see a low automation percentage. You might need to work with your developer to standardize the automation types used in the event data.

Check attributes

Go to the Case Attributes tab on the Process Insights page, and check the following attribute details to validate which cases are included in your process.

Expected attribute values

Do you see the attribute names and values you would expect for the process?

For example, you might know that the process involves global operations, so you would expect the process to include data for all geographical regions. If you see only cases with a Region attribute of APAC, work with your data steward and developer to determine whether the process is using the correct record types as data source or whether the process filters are configured incorrectly.

Also, are the attribute names confusing in any way? If so, you can rename the attributes, so that business users across your organization can easily understand them.

Empty attribute values

Do any attribute values appear as <No Value>?

This value is present if some cases in the process have a value for a specific attribute, but other cases do not.

The attribute value might be missing from certain cases if your organization started collecting data for the attribute after previously collecting case data without the attribute. If possible, work with your data steward and developer to identify a way to collect the historical data and add it to the older cases. If the historical data is unavailable, you can set a time period to exclude the historical cases with the missing values when you're exploring the analysis of that specific attribute.

However, if there is no historical difference in data collection, missing attribute values might indicate a misconfiguration in the source application or the process filters, so you'll want to work with your data steward and developer to troubleshoot potential issues.

Case count per attribute value

Do the case counts associated with specific attribute values match your expectations?

For example, you might see that the count of cases with a Country attribute of United Kingdom is significantly higher than the count of cases with a Country attribute of Spain. However, you're aware that the process is primarily performed by your company's office in Spain, so you should work with a data steward or developer to make sure the process includes all data from the Spain office.

Check activities and sequences

Check the following activity and sequence details to validate which cases are included in your process.

Expected activities

Go to the All Cases view of the process, then go to the Conformance and Costs tab.

Do you see all the activities you'd expect to see? Do you see activities you wouldn't expect to see?

For example, you might see activities associated with ad hoc events, which your data steward can remove from the process. Alternatively, you might see variations on what you know is a single type of activity, which your data steward can address by standardizing activities.

Activity order

One of Process HQ's greatest strengths is showing you the order in which activities actually occur in your process, so you can quickly remediate detours and delays. But, before you go all-in on analysis, you need to make sure that you're paying attention to valid differences in activity order and not differences that are artificially created by data quality issues.

Go to the Process Diagram tab of the Process Insights page. From start to end, the activities appear as nodes in an order determined by event timestamps.

Do the activities appear in the order you'd expect to see? Are the start activity and the end activity the ones you'd expect?

If the differences between your expectations and the activities you're seeing are extreme or suspicious, you might want to work with your data steward and developer to troubleshoot process preparation and filters. For example, you might want to look at standardizing timestamps in your event data if you haven't already done that.

Repetitions and loops

Go to the Process Diagram tab on the Process Insights page, and check for repetitions and loops.

By default, the Basic version of the process diagram is active, which displays the most common sequences in the most common order. You might want to change the complexity level to see how many alternate sequences Process HQ has identified from your data.

Are there reasonable explanations for those repetitions and loops, or could there be a problem with your data quality?

For example, do you see any repetitions in which the sequence connector shows a zero duration? This indicates that your data contains events with identical timestamps, which could be duplicates.

To confirm the presence of duplicate events:

1. In the process diagram, click the connector for the sequence that's not showing duration information. The sequence details dialog displays.
2. Go to the CASES tab.
3. Go to the Activity Details section for an individual case.

4. Review the timestamps for repeated activities in the list.

   For example, the image below highlights events with duplicate timestamps:

   

   Scroll or drag to pan · ⌘+/⌘− to zoom · Esc to close

Events with identical timestamps might be acceptable if other attribute values cause the events to differ from each other. For the most part, however, the presence of events with identical timestamps signals a problem with data quality, so you'll want to work with your data steward and developer to troubleshoot process preparation and filters.

Activity counts

Go to the All Cases view of the process, then go to the Conformance and Costs tab.

Compare the values in the Activity Count column to the overall case count:

• Activity count less than case count: Indicates that the activity does not occur in all cases in the process. Is that expected for this activity, or could it indicate incomplete data?

  Tip:  If the count of a terminal activity (for example, Account Funded) is less than the case count, it might simply indicate that a number of cases in your process are still in progress. If that is true, we recommend creating a view that shows only completed cases and exploring from there.

• Activity count more than case count: Indicates that the activity occurs more than once per case in the process. Is that expected for this activity?

  For example, your process might include an activity to create a customer account. A customer account can only be created once. If the Create Account activity occurs more than once per case, there might be duplicate events in your process data. Work with your data steward and developer to identify and eliminate duplicate events.

Activity and sequence duration

When you explore a process, you'll look even closer at activity and sequence durations. For now, though, your goal is simply to identify possible data quality issues by checking for extremes in activity and sequence durations.

Go to the Process Diagram tab of the Process Insights page. Next to each node and connector, you'll see the average duration Process HQ has calculated for that activity or sequence.

Are any durations too small? Do they appear as zero or another suspiciously low number? Or are any durations too big? Are they showing in terms of months or years when they should take only hours or days?

If you're seeing extremes of activity and sequence duration, work with your data steward and developer to make sure data collection is complete and appropriately prepared.

Feedback