A recent paper by Wagner and Kleiss compares a Ceilometer with Total Sky Imager and Micropulse lidar looking at suitability of ceilometers for  estimation of cloud amount.

https://journals.ametsoc.org/doi/pdf/10.1175/JTECH-D-15-0258.1

In summary they say,  in part:

“Ceilometers will be a mainstay of the operational automated weather observation network for years to come.

They are relatively inexpensive and require little ongoing maintenance, and are capable of 24-h observations.
While TSIs produce automated observations that are qualitatively more similar to human observations of sky cover than the ceilometer, ongoing operational
costs like mirror cleaning and reduced hours of operation limit this instrument’s applicability for deployment in unattended environments.

Two significant sources of error are associated with the automated sky cover observations obtained by single ceilometer ASOS installations: the spot view of the instrument renders it unable to see the entire sky, while the 3660-m height limit renders the instrument incapable of observing high clouds. These errors are not insignificant, and their magnitudes vary depending on the actual cloud coverage. The spatiotemporal averaging error is smallest for clear and overcast conditions as the sky exhibits little variability in these conditions. The high cloud error is at its smallest when skies are dominated by low clouds, and it tends to increase as low cloud coverage lessens, allowing high clouds to peek through.”

The ability to reliably detect upper atmosphere cloud is restricted to a subset of available ceilometers which have ranges in excess of the usual 7,500 to 10,000 metre range. Such ceilometers include the a CL51, the 8600-CHS and the CHM15K