Data that is received after the time it was expected or after related processing has already occurred.
Late-arriving data commonly refers to data that reaches a system after the time it was expected, or after downstream processing, reporting, alerting, or transaction posting has already taken place. The issue is about timing rather than whether the data is valid. The data may still be accurate and complete, but it arrives too late to be used in the intended sequence.
In manufacturing and regulated operations, late-arriving data can appear when machine events, inspection results, operator entries, supplier confirmations, or integration messages are delayed between OT and IT systems such as PLCs, SCADA, MES, LIMS, QMS, ERP, or analytics platforms. For example, a production count may post after a shift report is closed, or a quality result may arrive after a lot has already advanced to the next step.
Late-arriving data does not automatically mean bad data, missing data, or duplicate data. It is different from incorrect timestamps, although timestamp errors can make late arrival harder to detect. It is also different from real-time data loss, where the data never arrives at all.
Operationally, late-arriving data matters because many manufacturing workflows depend on event order and timing. Delayed records can affect traceability, KPI calculations, exception handling, inventory status, quality holds, electronic records, and system-to-system reconciliation. Systems often need logic to decide whether to accept, reprocess, flag, or version a late record so that history remains understandable.
Late-arriving data is often confused with stale data and backdated data. Stale data is old data that has not been refreshed. Backdated data is entered with an earlier effective date, which may or may not have arrived late. A late-arriving record can also be backdated, but the terms are not identical.