Service incident - 19th of June 2026 - Guess correction incorrectly applied to fill-in-the-gap questions with hidden gaps

All times in this document are recorded in UTC+2 (CEST).

Article update – July 8th, 2026

Following publication of this incident report, we identified an additional scenario that was not included in the original impact analysis. In addition to questions containing hidden (alpha)numeric gaps, the issue could also occur for fill-in-the-gap questions that contained only dropdown gaps, where one or more dropdown gaps had been removed and remained hidden in the question configuration.

Although these hidden dropdown gaps were not visible to students and could not be answered, they were still included in the guess score calculation, causing the calculated guess score to be higher than intended.

The impact analysis was repeated to include this scenario. As a result, the total impact increased to 2,551 results across 31 assignments. All affected schools have been informed.

Incident summary

An issue was identified where hidden gaps incorrectly influenced the guess correction of certain fill-in-the-gap questions.

Although these hidden gaps were not visible to students and could not be answered, they remained present in the underlying question configuration. Depending on the question configuration, this could either prevent guess correction from being applied altogether or cause the calculated guess score to be higher than intended.

As a result, participants may have received incorrect scores and marks.

A fix for this issue was deployed on June 19, 2026, at 13:39.

Following publication of the original incident report, an additional affected scenario was identified during a follow-up review. The impact analysis was repeated to include this scenario.

The updated investigation identified 2,551 impacted results across 31 assignments.

The affected schools have been informed.

Lead-up

On June 17, 2026, at 13:53, a user contacted Support after noticing that guess correction was not being calculated for a fill-in-the-gap question.

The reported behaviour was investigated by the technical team. Initial analysis showed that certain questions containing only dropdown gaps did not receive guess correction despite meeting the requirements for guess correction to be applied.

Further investigation revealed that hidden gaps contained within the question configuration influenced the guess score calculation.

Following deployment of the initial fix and completion of the first impact analysis, a further review of the implementation identified an additional scenario in which hidden gaps could also incorrectly affect the calculated guess score. The impact analysis was subsequently repeated to include this scenario.

Fault

This issue has existed since the construction page of fill-in-the-gap questions was updated on December 7, 2025. As part of that update, removed gaps became restorable. Rather than being immediately deleted, removed gaps were retained in a hidden state within the question configuration until they were permanently deleted by a user.

Fill-in-the-gap questions can contain multiple types of gaps, including dropdown gaps and (alpha)numeric gaps. When gaps are removed from a question, they remain present in the underlying question configuration while no longer being visible to students or answerable. These gaps can be restored until they are permanently deleted.

The guess score calculation incorrectly considered these hidden gaps when determining how guess correction should be applied. Since hidden gaps cannot be answered by students, they should not influence the calculation.

This resulted in two affected scenarios:

• If a hidden (alpha)numeric gap remained in the question configuration, the question could be incorrectly classified as containing (alpha)numeric gaps. Because guess correction is not applied to questions containing answerable (alpha)numeric gaps, guess correction was incorrectly skipped entirely.
• If a hidden dropdown gap remained in the question configuration, it incorrectly increased the number of possible answer combinations considered during the guess score calculation. This caused the calculated guess score to be higher than intended.

In both cases, the guess correction no longer reflected the question that was actually presented to students.

Impact

The impact of this incident could have affected student scores and marks.

Following publication of the original incident report, an additional affected scenario was identified and a complete impact analysis was repeated.

The updated investigation identified 2,551 impacted results across 31 assignments.

All affected schools, including those newly identified by the updated investigation, have been informed and provided with an overview of the impacted assignments and results.

Detection

The issue was reported by a user via a support ticket. The technical team investigated the reported behaviour and identified the underlying cause.

Following deployment of the initial fix, an additional review of the implementation identified a second scenario that was not included in the original investigation. This prompted a complete re-analysis of the impact.

Response

The Support team forwarded the issue on June 17, 2026, at 14:55 to the technical team.

The technical team started investigating the issue on June 18, 2026, at 15:04. A fix was developed, reviewed and deployed on June 19, 2026, at 13:39.

Schools with assignments scheduled for publication were informed as soon as impacted assignments were identified.

After deployment, an impact analysis was performed and the affected schools were informed. Once the additional affected scenario was discovered during a follow-up review, the impact analysis was repeated in full and any newly affected schools were informed.

Recovery

A fix was deployed on June 19, 2026, at 13:39 to ensure that hidden gaps no longer influence the guess score calculation of fill-in-the-gap questions.

Following deployment, newly calculated guess scores correctly ignore hidden gaps regardless of their type.

Timeline

17th of June, 2026

• 13:53 - The Support team received a ticket reporting that guess correction was not applied to a fill-in-the-gap question.
• 14:55 - The Support team created a task for the technical team to investigate the issue.

18th of June, 2026

• 15:04 - The technical team started investigating the issue.

19th of June, 2026

• 10:49 - The technical team identified an impacted assignment scheduled for publication that day.
• 11:25 - The Support team informed the affected institution.
• 12:00 - The initial impact analysis was completed.
• 13:39 - The hotfix was deployed to the production environment.
• 15:58 - The initial impact was confirmed.
• 16:35 - The initially affected schools were informed.

30th of June, 2026

• 10:02 - An additional affected scenario was identified during a follow-up review.

8th of July, 2026

• 09:25 - The final impact analysis was completed and the complete scope of the incident was confirmed. 
• 10:15 - All affected schools were informed. 

Reflection

This issue was introduced on December 7, 2025, when the construction page for fill-in-the-gap questions was updated to support restoring removed gaps.

The root cause was that the guess score calculation did not consistently distinguish between answerable gaps and hidden gaps. Hidden gaps remained part of the internal question configuration and were incorrectly considered during guess score calculation even though students could not interact with them.

The initial investigation correctly identified one manifestation of the issue, where hidden (alpha)numeric gaps prevented guess correction from being applied. However, after the incident had been closed, a follow-up review identified a second manifestation: hidden dropdown gaps could also incorrectly increase the number of possible answer combinations used during guess score calculation. This scenario was not included in the initial impact analysis, resulting in additional affected assignments and results being identified after publication of the original incident report.

After discovering this additional scenario, the impact analysis was repeated in full and all newly affected schools were informed. This incident report has been updated to reflect the complete scope of the issue.

This incident demonstrates the importance of validating scoring logic against every possible hidden question state rather than focusing only on the most obvious manifestations of a defect. Future changes affecting question configuration and guess score calculations will receive additional regression testing to ensure that hidden configuration elements cannot influence participant scores.