Context : The well established negative correlation between the \alpha _ { OX } spectral slope and the optical/UV luminosity , a by product of the relation between X-rays and optical/UV luminosity , is affected by a relatively large dispersion .
The main contributions can be variability in the X-ray/UV ratio and/or changes in fundamental physical parameters .

Aims : We want to quantify the contribution of variability within single sources ( intra-source dispersion ) and that due to variations of other quantities different from source to source ( inter-source dispersion ) .

Methods : We use archival data from the XMM-Newton Serendipitous Source Catalog ( XMMSSC ) and from the XMM-OM Serendipitous Ultra-violet Source Survey ( XMMOM-SUSS3 ) .
We select a sub-sample in order to decrease the dispersion of the relation due to the presence of Radio-Loud and Broad Absorption Line objects , and to absorptions in both X-ray and optical/UV bands .
We use the Structure Function ( SF ) to estimate the contribution of variability to the dispersion .
We analyse the dependence of the residuals of the relation on various physical parameters in order to characterise the inter-source dispersion .

Results : We find a total dispersion of \sigma \sim 0.12 and we find that intrinsic variability contributes for 56 \% of the variance of the \alpha _ { OX } - L _ { UV } relation .
If we select only sources with a larger number of observational epochs ( \geq 3 ) the dispersion of the relation decreases by approximately 15 % .
We find weak but significant dependences of the residuals of the relation on black-hole mass and on Eddington ratio , which are also confirmed by a multivariate regression analysis of \alpha _ { OX } as a function of UV luminosity and black-hole mass and/or Eddington ratio .
We find a weak positive correlation of both the \alpha _ { OX } index and the residuals of the \alpha _ { OX } - L _ { UV } relation with inclination indicators , such as the FWHM ( H \beta ) and the EW [ O _ { III } ] , suggesting a weak increase of X-ray/UV ratio with the viewing angle .
This suggests the development of new viewing angle indicators possibly applicable at higher redshifts .
Moreover , our results suggest the possibility of selecting a sample of objects , based on their viewing angle and/or black-hole mass and Eddington ratio , for which the \alpha _ { OX } - L _ { UV } relation is as tight as possible , in light of the use of the optical/UV - X-ray luminosity relation to build a distance modulus ( DM ) - z plane and estimate cosmological parameters .

Conclusions :