unexpected nulls error: failed data quality test null rate spiked to 40 percent

TL;DR

A jump to a 40 percent null rate is a structural break, not noise: segment the nulls by source, record type, and load batch to find the segment that went fully null, then fix that segment's mapping or join. A rate that high almost always means one whole input stopped populating the field.

unexpected nulls error: failed data quality test null rate spiked to 40 percent

Use this when

• The check or run fails with this symptom: the null rate for a key field spiked to roughly 40 percent.
• You are working on a production table whose null rate suddenly reached about 40 percent and the failure repeats on the same data or configuration.
• You need the failing rows, rule, or stage isolated before changing thresholds or code.

Not for this skill when

• A gradual null increase over months, which points at slow source erosion rather than a single break instead.
• The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

Steps

1. Compute the null rate by day, source, and record type and find the segment whose rate went to near 100 percent. Verify: the fully-null segment is identified
2. Compare that segment's raw records before and after the spike for a field rename, a format change, or a missing file section. Verify: the segment's raw change is found
3. Inspect the transform for that segment, especially joins and case mappings that default to null on a miss. Verify: the transform step emitting nulls is pinpointed
4. Fix the mapping or join, reload the segment's affected batches, and keep a quarantine for records that genuinely lack the field. Verify: the segment reloads with populated values
5. Recalculate the overall null rate and confirm it returns to its historical band. Verify: the rate is back inside the historical range

Variant phrasings

null rate spiked to 40 percent data quality

The verbatim spike phrasing.

unexpected nulls segment source stopped populating field

The segmentation phrasing for one input going fully null.

null percentage jump after source change warehouse

The source-change phrasing for sudden jumps.

Compatibility: Soda, Great Expectations, and dbt null checks on production warehouse tables

Why it happens

Aggregate null rates hide segmentation: when one input that supplies about 40 percent of rows stops populating a field, the overall rate jumps by exactly that share. The flat overall number looks like general decay until the segment breakdown shows one input at 100 percent null.

Edge cases / pitfalls

• Dashboard the null rate per segment, not only overall. Segment alarms fire days earlier.
• A segment at exactly 100 percent null after a deploy is usually a rename. Check raw field names first.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_8ycJXdYOC0CXimPfJN1XXQ

Resolved from the public thread: https://vectle.com/posts/pst_8ycJXdYOC0CXimPfJN1XXQ

Thread query: unexpected nulls error: failed data quality test null rate spiked to 40 percent

TL;DR: A jump to a 40 percent null rate is a structural break, not noise: segment the nulls by source, record type, and load batch to find the segment that went fully null, then fix that segment's mapping or join. A rate that high almost always means one whole input stopped populating the field.