No dataset is perfect - and Network Rail’s JSON is no exception.
Here are some highlights from the QA perspective.
It’s still the early days of the Level Crossing Database as a project.
The focus for now is to explore the data and make it accessible, rather than improve it.
In the short term, all feedback is welcome via GitHub issues: you can just sign in using a Google account if you don't have one on GitHub.
Several options are in play in the medium term. The simplest would be an AppSheet app where users can report issues for individual level crossings, and these are added to a Google Sheet for review. This would inform the next steps to be taken.
And above all, the project welcomes collaboration. Are you a stakeholder with an interest in this data? Or are you an enthusiast with time and skills to help develop the database and the looker studio? Get in touch.
The source JSON doesn’t appear to be up to date. One of the fields is Next Assessment Date, and as of July 2025, the 'Next Assessment Date' for all crossings is in the past.
Numerous crossings have blank or invalid values for one field or another. Missing Position is the most problematic, and providing a process for fixing this is currently a priority. The best way to stay up to date for now is to visit the GitHub and "watch" the repository with "all activity" selected.
Issues with consistency aren’t always obvious, but two examples are easily detected:
The line speed unit used for most crossings is assumed to be mph, but one crossing on the WCML shows a line speed of 200 which suggests it is kph.
There are many crossings of two types, generally a Public Footpath Crossing and a Private User worked Crossing. This is often reflected as two crossings in one Position, but sometimes these Positions are captured with different names, while at other times they are identical.
Above all, no official documentation for this dataset has been found so far. It would clearly be preferable to link to any documentation that might be in the public domain and any such links would be appreciated. You can submit one via GitHub - or just get in touch.