JSON Drift Brief

Compare two JSON samples before a downstream workflow consumes changed data. See nested fields that appeared or disappeared, changes in value types, and shifts in missing, null, blank or empty-container rates. Download a readable HTML brief and a JSON report.

Try the sample

Keep the prefilled Before and After samples, then run. Open Readable HTML report or Full JSON report. The example reports four signals: rating removed, price changed from number to string, promo added, and blank titles rising from 0% to 40%.

Paste your own JSON arrays of objects to compare data. URLs and file paths are not fetched. No scraper credentials are required. This is an explicit comparison tool, not a scraper or continuous monitoring service. It does not claim to be the only drift tool or replace schema validation.

How it works

• Object fields use JSON Pointer paths, e.g. /details/stock. Literal / and ~ in keys are escaped.
• Field additions/removals and changes in the observed type set always appear.
• Presence, null, blank string and empty-container rates are separate; rates use all rows in the respective sample as denominator. Default threshold: 10 percentage points.
• Zero and false are not empty. Empty samples are inconclusive.
• No reported signals means no changes in inspected properties at the chosen threshold, not proof the data is correct. Different row counts are displayed; this is not a significance test.

Limits

Maximum 2,000,000 UTF-8 bytes and 20,000 rows per sample, 32 nesting levels below the root array, 250,000 nodes, and 5,000 distinct paths. Each field path is limited to 2,048 UTF-8 bytes, with at most 2,000,000 combined bytes of distinct paths per sample. These path limits also bound report size. Exceeding limits fails clearly.

Array elements are not inspected. Array fields and empty-array rates are checked; paths are disclosed. Changes inside arrays may be missed. No row identity matching, numeric distribution checks, individual value comparison or semantic validation. Small or unrepresentative samples may mislead.

Duplicate keys, malformed JSON, non-finite numbers and invalid Unicode are rejected. Source row values are not written to application logs or reports. Field names are included and may be sensitive.

Data handling

This Actor runs on Apify, not in your browser. Input is sent to Apify and stored with the run; output is stored in the run's dataset and key-value store. Access and retention follow your Apify settings and policy. Do not paste secrets or data you are not authorized to process. The application sends no data to external APIs and fetches no external URLs.

Pricing and beta status

Experimental price: US$0.03 per completed comparison report, including inconclusive reports from empty samples. This is one charge per report, not per input row, field or detected change. Check the displayed Apify pricing panel before running.

The report price includes processing. There is no separate startup charge. A completed comparison includes both the HTML and JSON reports.

Set the run budget to at least the displayed price for one report. If the remaining budget is insufficient, the run stops before creating a report. Rejected input and insufficient-budget runs do not incur the comparison-report charge. Apify's displayed terms govern any other applicable platform charges.

This is an early product with the limits described above. Please check its example and supported comparisons before using it in a production workflow.

Development

Python 3.13+, pip install -r requirements.txt, python -m unittest discover -s tests -v. To test the Actor locally, copy examples/INPUT.json to storage/key_value_stores/default/INPUT.json, then python -m src, without cloud credentials.