Lever Says a Data Pipeline Stopped Updating Its Analytics Reports, and the Issue Was Identified After Customers Noticed Their Report Counts Did Not Match the Live Data

Lever posted a postmortem on its status page on June 22, 2026, under the title "Stale Visual Insights Data." Between June 14 and June 18, 2026, the postmortem says, analytics reports in Visual Insights "displayed outdated information." Parts of the postmortem and of the updates posted during the incident follow, not the whole.

Accurate Platform, Outdated Reports

"While the live platform remained fully accurate and no data was lost, the reports stopped reflecting real-time changes (such as candidate stage updates)." The result, the postmortem says, was an outdated snapshot of data shown to "all reporting customers for approximately four days."

The issue, it says, "was identified on June 17, 2026, following reports from customers who noticed that their report counts did not match the live data in the platform." The engineering team then confirmed "that the data pipeline responsible for updating these reports had stopped processing new information on June 14."

Two Start Dates and a Fix

The first update on the status page, posted on June 16, reported an issue affecting the North America data center, where Visual Insights was not loading new data in some customer accounts, and said the issue "started occurring on Tuesday, 6/16/2026 at approximately 7:30am PST/PDT." A later update that day said, "New data may be delayed to appear in Visual Insights or may be missing."

The postmortem, posted after the fix, says the pipeline had stopped processing new information on June 14. The status page carries both dates.

A fix was implemented on June 17 "at approximately 10:30am PST/PDT," according to the update posted that morning. In the postmortem's words, the engineering team "restarted the affected component, which safely resumed and caught up on all pending changes with no data loss." The update of June 18 said the data backfill "has completed at approximately 5:00am PST/PDT on 6/18/2026." No action, the postmortem says, "was required from customers at any point."

Three Measures

The postmortem lists three measures. The system had been updated "to automatically recognize and recover from this specific type of database restart error, allowing it to self-heal without manual intervention." More granular alerts were being implemented "that monitor the specific workers within the data pipeline," so that "if a single worker stops, we are notified within minutes." And new automated checks were being added "that compare live platform data against the reporting store," to detect gaps in freshness immediately.