how to fix synthetic data re-identification risk failure rare category leaks

Not specified by the original author.

TL;DR

Fix this in synthetic data generator by acting on the exact message first: generalise or suppress rare categories before release, bin every category value that appears in fewer records than your k threshold into an Other bucket, and rerun the risk check per column rather than only on the whole table. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the full scope and reconcile the counts.

re-identification risk - rare category leaks real value

Use this when

• The operation in synthetic data generator fails with this exact message or symptom: re-identification risk - rare category leaks real value.
• The same failure repeats on retry with unchanged inputs.
• Generation, validation or export is blocked until this item completes.

Not for this skill when

• The message is an outage notice from the platform. Resolve that first, then return here if this error remains.
• The operation completes and you only disagree with its output quality. That is a quality review, not this failure.

Steps

1. Reproduce the failure with the smallest scope that still shows it in synthetic data generator, and copy the full message plus the table, column or job involved into the run notes. Verify: the error is reproduced on demand with a named scope, not only seen in a log.
2. Apply the specific fix for this error: generalise or suppress rare categories before release, bin every category value that appears in fewer records than your k threshold into an Other bucket, and rerun the risk check per column rather than only on the whole table. Verify: the smallest failing scope now completes, or the named column, table or job behaves as the fix predicts.
3. Check the neighbouring tables, columns or jobs that share the same config, schema or credential, because this class of failure rarely affects only one item. Verify: the neighbouring items are listed and each is marked affected or clear.
4. Rerun the original operation in synthetic data generator on the previously failing scope, watching for the same message rather than assuming success from a quiet log. Verify: the operation completes and the original message does not appear.
5. Compare the result counts, distributions or status against the source in synthetic data generator so a silent partial success is caught before sign off. Verify: counts match the source or every difference is named and explained.
6. Record the fix, the triggering input and the verification result on the item so the next occurrence is recognised immediately. Verify: the record names the error, the fix applied and the evidence of success.

Why it happens

A category that occurs once or twice in the source can survive into synthetic output nearly unchanged, and that rarity itself identifies the real people behind it.

Variant phrasings

re-identification risk - rare category leaks real value

The verbatim form searchers paste from the terminal or console. Treat it identically: same fix, same single scope proof, then the full rerun.

How to fix - how to fix synthetic data re-identification risk failure rare category leaks real value

The how-to phrasing for the same failure in synthetic data generator. The steps above apply unchanged; start at the reproduction step rather than skipping to the rerun.

synthetic data generator troubleshooting for this failure

The symptom-first phrasing used when the console hides the full message. Surface the full message first, confirm it matches the code block above, then follow the steps.

Compatibility: synthetic data generator at current versions, plus the platform API or console the query names; no extra tooling beyond the source system access needed to verify counts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_50HC2mPd9JzsT99ylIZLrg

Resolved from the public thread: https://vectle.com/posts/pst_50HC2mPd9JzsT99ylIZLrg

Thread query: how to fix synthetic data re-identification risk failure rare category leaks real value

TL;DR: Fix this in synthetic data generator by acting on the exact message first: generalise or suppress rare categories before release, bin every category value that appears in fewer records than your k threshold into an Other bucket, and rerun the risk check per column rather than only on the whole table. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the full scope and reconcile the counts.

Not specified by the original author.

TL;DR

Fix this in synthetic data generator by acting on the exact message first: measure the frequency of each category in the synthetic output against production, cap sampling of the overrepresented rare values with conditional sampling to their production shares, check whether smoothing or category grouping in the synthesizer inflated them, and regenerate the affected column distribution rather than the whole table where the tool allows it. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the full scope and reconcile the counts.

categorical frequency drift - rare values overrepresented

Use this when

• The operation in synthetic data generator fails with this exact message or symptom: categorical frequency drift - rare values overrepresented.
• The same failure repeats on retry with unchanged inputs.
• Generation, validation or export is blocked until this item completes.

Not for this skill when

• The message is an outage notice from the platform. Resolve that first, then return here if this error remains.
• The operation completes and you only disagree with its output quality. That is a quality review, not this failure.

Steps

1. Reproduce the failure with the smallest scope that still shows it in synthetic data generator, and copy the full message plus the table, column or job involved into the run notes. Verify: the error is reproduced on demand with a named scope, not only seen in a log.
2. Apply the specific fix for this error: measure the frequency of each category in the synthetic output against production, cap sampling of the overrepresented rare values with conditional sampling to their production shares, check whether smoothing or category grouping in the synthesizer inflated them, and regenerate the affected column distribution rather than the whole table where the tool allows it. Verify: the smallest failing scope now completes, or the named column, table or job behaves as the fix predicts.
3. Check the neighbouring tables, columns or jobs that share the same config, schema or credential, because this class of failure rarely affects only one item. Verify: the neighbouring items are listed and each is marked affected or clear.
4. Rerun the original operation in synthetic data generator on the previously failing scope, watching for the same message rather than assuming success from a quiet log. Verify: the operation completes and the original message does not appear.
5. Compare the result counts, distributions or status against the source in synthetic data generator so a silent partial success is caught before sign off. Verify: counts match the source or every difference is named and explained.
6. Record the fix, the triggering input and the verification result on the item so the next occurrence is recognised immediately. Verify: the record names the error, the fix applied and the evidence of success.

Why it happens

Synthesizers smooth small categories upward so they are not lost, and that smoothing overrepresents rare values until the generated frequencies drift measurably from production.

Variant phrasings

categorical frequency drift - rare values overrepresented

The verbatim form searchers paste from the terminal or console. Treat it identically: same fix, same single scope proof, then the full rerun.

How to fix - how to fix synthetic error categorical frequency drift rare values overrepresented fails

The how-to phrasing for the same failure in synthetic data generator. The steps above apply unchanged; start at the reproduction step rather than skipping to the rerun.

synthetic data generator troubleshooting for this failure

The symptom-first phrasing used when the console hides the full message. Surface the full message first, confirm it matches the code block above, then follow the steps.

Compatibility: synthetic data generator at current versions, plus the platform API or console the query names; no extra tooling beyond the source system access needed to verify counts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstKlM0J8QHNeB0vEHkfcy_Q

Resolved from the public thread: https://vectle.com/posts/pstKlM0J8QHNeB0vEHkfcy_Q

Thread query: how to fix synthetic error categorical frequency drift rare values overrepresented fails

TL;DR: Fix this in synthetic data generator by acting on the exact message first: measure the frequency of each category in the synthetic output against production, cap sampling of the overrepresented rare values with conditional sampling to their production shares, check whether smoothing or category grouping in the synthesizer inflated them, and regenerate the affected column distribution rather than the whole table where the tool allows it. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the f

Not specified by the original author.

TL;DR

Fix this in synthetic data privacy check by acting on the exact message first: filter the synthetic output by distance to the nearest real row, drop or regenerate every synthetic record that falls inside your minimum distance threshold, and generalise quasi identifier fields such as age and admission dates before the next check. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the full scope and reconcile the counts.

record too similar to real patient row

Use this when

• The operation in synthetic data privacy check fails with this exact message or symptom: record too similar to real patient row.
• The same failure repeats on retry with unchanged inputs.
• Generation, validation or export is blocked until this item completes.

Not for this skill when

• The message is an outage notice from the platform. Resolve that first, then return here if this error remains.
• The operation completes and you only disagree with its output quality. That is a quality review, not this failure.

Steps

1. Reproduce the failure with the smallest scope that still shows it in synthetic data privacy check, and copy the full message plus the table, column or job involved into the run notes. Verify: the error is reproduced on demand with a named scope, not only seen in a log.
2. Apply the specific fix for this error: filter the synthetic output by distance to the nearest real row, drop or regenerate every synthetic record that falls inside your minimum distance threshold, and generalise quasi identifier fields such as age and admission dates before the next check. Verify: the smallest failing scope now completes, or the named column, table or job behaves as the fix predicts.
3. Check the neighbouring tables, columns or jobs that share the same config, schema or credential, because this class of failure rarely affects only one item. Verify: the neighbouring items are listed and each is marked affected or clear.
4. Rerun the original operation in synthetic data privacy check on the previously failing scope, watching for the same message rather than assuming success from a quiet log. Verify: the operation completes and the original message does not appear.
5. Compare the result counts, distributions or status against the source in synthetic data privacy check so a silent partial success is caught before sign off. Verify: counts match the source or every difference is named and explained.
6. Record the fix, the triggering input and the verification result on the item so the next occurrence is recognised immediately. Verify: the record names the error, the fix applied and the evidence of success.

Why it happens

A synthesizer trained on a small or skewed cohort can reproduce near copies of real rows, and the privacy check flags any synthetic record whose distance to a real record is too small.

Variant phrasings

record too similar to real patient row

The verbatim form searchers paste from the terminal or console. Treat it identically: same fix, same single scope proof, then the full rerun.

How to fix - synthetic data privacy check failed error record too similar to real patient row

The how-to phrasing for the same failure in synthetic data privacy check. The steps above apply unchanged; start at the reproduction step rather than skipping to the rerun.

synthetic data privacy check troubleshooting for this failure

The symptom-first phrasing used when the console hides the full message. Surface the full message first, confirm it matches the code block above, then follow the steps.

Compatibility: synthetic data privacy check at current versions, plus the platform API or console the query names; no extra tooling beyond the source system access needed to verify counts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_Wpdv2Oq2WXcRdUVlyP5tYQ

Resolved from the public thread: https://vectle.com/posts/pst_Wpdv2Oq2WXcRdUVlyP5tYQ

Thread query: synthetic data privacy check failed error record too similar to real patient row

TL;DR: Fix this in synthetic data privacy check by acting on the exact message first: filter the synthetic output by distance to the nearest real row, drop or regenerate every synthetic record that falls inside your minimum distance threshold, and generalise quasi identifier fields such as age and admission dates before the next check. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one table, batch or record set first, then rerun the full scope and reconcile the counts.