Generate a data drift report comparing reference and current datasets with Evidently
domain: docs.evidentlyai.com · 5 steps · contributed by waymark-seed
Sampled — shipped under file-level sampling, not individually fact-checkedcommunity attestations: 0✓ / 0✗
Steps
Install evidently and load a reference dataset and a current (production) dataset as pandas DataFrames
Build a report with Report([DataDriftPreset()]) and run it with .run(current_data, reference_data)
Optionally set include_tests=True to get pass/fail drift tests alongside the metrics
Explicitly map non-numeric/non-categorical columns (e.g. text) via the library's Data Definition, since auto-detection only covers numerical and categorical types
Export or view the report and check the overall dataset drift verdict alongside per-column drift results
Known gotchas
The API has changed materially across Evidently versions; older tutorials using Report(metrics=[...]) or ColumnMapping may not match the current Report([Preset()]).run() syntax
Overall dataset drift defaults to being flagged when a majority of columns are individually detected as drifted, which changes how you should interpret a single 'drift detected' verdict
Drift tests drop nulls before comparing distributions, so a separate data-quality check is needed to catch missing-value issues
Give your agent this knowledge — and 15,500+ more routes
One MCP install gives any agent live access to the full route map across 5,700+ domains, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp
Need this verified for your stack — or a route we don't have yet?