Data 360: Troubleshooting Segmentation Errors

What to do if there are count failures?

In cases of count failures, you’ll find mention of the ‘count’ word in the error. Let's understand what you need to check when you encounter count failures. 
 

• In such cases, try to simplify the segment query further using the following methods:
  1. Merge containers wherever possible - Identify containers that have the same join path, and merge them. Use "Nested Operators" instead to create the correct filter logic. This will simplify your segment. Please reach out to Support if you do not see nested operators in your org.
  2. Long join paths - In a segment, while selecting related attributes, always and absolutely, as a best practice, use the shortest path to reach the right DMO as long as it doesn’t affect the grain. For example: if a join path is a→b→c→d→e and there’s a shorter path like a→c→e, then check the join relationship in DMOs and if the grain isn’t compromised, try to use a shorter path
  3. Cyclic Paths - Most often, your count queries will time out or fail, or the status will remain stuck if you have used cyclic paths in your segments. 
     1. What are cyclic paths - simply put, a→b→c→a , or a → b → c → b
     2. How to resolve them - So far, we have not encountered any scenario where cyclic paths were unavoidable. Once you have identified the cyclic path, check what alternate exists to reach the same results. For example, in this case, a→b→c→a, you can look for a direct attribute within a DMO; or in a → b → c → b, you can look for an attribute in b, rather than fetching it from c.

Try again later

• You’ll mostly encounter this issue when the resources allotted to a tenant are exhausted. Simplifying the segment will help here. 
  1. If the error is “Multiple counts in progress, try again later.” - When you get this error, all you can do is try the count again later. Please note that it’s difficult to quantify the term “later” here as there could be other queries enqueued for processing. This “wait time” will vary from tenant to tenant.

What to do if there are materialization failures?

In case of materialization failures, most often than not, you’ll see “segment could not be processed” or “segment could not be published.” In this case, you’ll also see that simplifying a segment helps significantly. 

• Let’s break this down further and see what we can do to fix this issue. 
  1. If a segment is tagged as “Large” - This will mostly happen while working with Rapid segment. Any segment that has been tagged as “Large", the publish will fail. Please note that segment count too can fail if the segment is tagged as large during design time itself.
     • How to fix it - The easiest way to handle this situation is to reduce the number of processed records. As such, we auto-apply a filter of 7 days on engagement date of “Engagment DMO”. In some cases we have observed that segments were being tagged as “Large” because engagement type data wasn’t tagged to “Engagement DMO”. You see when this happens, the lookback period is not applied on the dmo data. Please ensure your DMO type is correct.
  2. Tenant not found -
     • You should not encounter this error, ideally. If you do, please create an investigation with Support and provide the error message. 

Other Issues

Data Mismatches

This is one situation you can get into many times. Let’s look at a couple of scenarios we have come across in the past:

1. Count in Data Explorer, and the segment does not match - 
   •  Segmentation applies a filter on the engagement date automatically - 24 months for standard segments and 7 days for rapid. Please ensure in data explorer you have applied the filter on the engagement event date as well
   •  Segment provides a count of unique profiles - whereas Data Explorer provides you with raw data available in the source.
2. Count in other tools is not matching
   • Please make sure you have checked point 1.
   • Date Operators - Date operators in segments consider tenant timezone. Please go through the docs very carefully to ensure the comparison is apples to apples.
   • Operator Behavior - Again, make sure the comparison is apples to apples, and understand the behavior of operators completely. For example: if you are checking for nulls, make sure you check what the definition of null in both tools is.
3. Drastic changes in Segment population
   • Check if the source data has gone through drastic changes - significant addition, updates, or deletions.
   • If you’re using Calculated Insights in your segment - If you're using Calculated Insights in your segments and notice an unexpected increase or decrease in segment count and/or segment population, please check if any Identity Resolution run was run between CI and segment run.

0 Excluded records in Segment population:

Excluded Records in Segment Population:When using an AND in a Segment Exclude, each bucket of criteria must have records that fall under the criteria. If each bucket of the AND statement does not have a record that meets the criteria, the Exclude is ignored altogether.

For example: If you have an Exclude condition as Condition1 AND Condition2 AND Condition3, please ensure that the record meets all three conditions. If any one of the conditions is not met, the record will not be excluded and will be included in the Segment if it satisfies the include criteria. If you want to exclude a record that meets any one of the conditions, use OR instead of AND, and this would exclude the record. 

Segment Status Mismatch

Sometimes, you’ll see that segment materialization is successful (successful publish history exists), but the segment throws an error. This is due to a sync mismatch. One of the easiest ways to get around this issue is to click on the “Update Status” button manually. The error should go away.

Could not load ALM graph

If you are facing issues with segment edit window timing out or unable to load the ALM graph, then the issue is mostly with the data model defined at DMO levels. To debug further, check the relationships defined on the “Segment On” DMO and see if relationships can be disabled. If the issue persists, create an investigation with Support.

Unable to deactivate segment

If you are trying to deactivate a segment and get an "Segment could not be deactivated. Please Try Again." error message, this may be due to the segment still containing an active or inactive publishing schedule. Simply edit the segment properties to set it to "Don't refresh" before attempting to deactivate the segment again.