DataPilot / Data quality checks
Know what’s wrong
with your data before you analyze it.
Automated checks on every signal and trip.
Each issue is shown on the signal, with the check that found it.
Know what’s wrong
TISAX certifiedEU-hostedOn-premise available
Checks
Every check.
On every mapped signal.
Thresholds are set automatically from the signals DataPilot detects – nothing to configure.
Stale State
The vehicle is moving, but the signal stops updating.
Null Values
Null or NaN values where the logger recorded nothing.
Offset Consistency
Related signals, like the four wheels, sit at different baselines.
Value Range
Samples the signal can’t physically reach.
Outliers
Non-physical spikes from sensor or electrical faults, not from driving.
Signal Jumps
One step far larger than the signal’s usual change.
Signal Rates
A whole segment changes faster than is physically plausible.
Noise
High-frequency jitter that hides real events.
Saturation
The sensor hits its physical ceiling and flat-lines. Peaks look lower than they were.
Standard Value
Values like 255 mean “no reading” – left in, they look like extreme peaks.
Signal Unit
The values fit another unit better than the one mapped.
Resolution
Steps between values are too large for the analysis.
Categorical Columns
A continuous signal behaves like a handful of categories.
Mileage Overlap
Duplicated recordings, or trips missing between files.
Monotonicity
A value that can only grow goes down.
Inspect
Every issue shows where and why.
Fix
Fixes you can see before you apply them.
How fixes work
- Only flagged samples change
- Raw data is kept. Fixes create a cleaned version.
- See the impact first: trips affected, % of samples
- All files and trips in one run
- Apply a fix to every signal and trip with the same issue
- PDF summary for your test report
Fix methods
Fill interpolation
Fills a gap linearly from neighboring values
Fill previous
Holds the last valid value
Clip bounds
Caps values at the physical min and max
Scale signal
Corrects gain and offset, e.g. a wrong unit
Drop signal
Removes the signal, keeps the rest of the trip
Drop trip
Removes the whole trip
Signal mapping
Signal names are mapped before the checks run.
Raw names are matched to standard DataPilot signals – you review and confirm each match. The Signal Unit check flags values that fit another unit, e.g. mbar instead of bar.
Roadmap
Available now · Coming next
Available now
- Automated checks on every signal
- Fix methods with impact preview
- One fix for every signal with the same issue
- Raw data kept, cleaned version exported
- PDF summary
- Export in MAT, CSV or Parquet
Coming next
- MF4 files
- Dismiss a flagged issue that is real signal behavior, not a fault – it no longer counts in the score
- Anomaly detection check
- Adjustable thresholds
FAQ
Questions about data quality checks
No. DataPilot runs every check automatically, with defaults based on the signals it detects.
Not yet. Adjustable thresholds are on the roadmap.
It stays untouched. Fixes create a cleaned version of your dataset, and that is the version you export. Fixes can’t be undone, but the raw data is always kept.
Leave it as it is. DataPilot never applies fixes automatically – you choose what to fix. Soon you’ll also be able to dismiss it, so it no longer counts in the score.
MAT, CSV and Parquet. MF4 support is coming soon.
Building the next generation of automotive data infrastructure. Certified TISAX | EU-Hosted | © 2026 COMPREDICT GmbH – Darmstadt, Germany
All Rights Reserved
