In a well-designed survey of 341 organizations, they identify significant correlations between IT performance and ITSM best practices, but also find some surprising non-correlations. For example, simple Change oversight does not predict performance in their sample, but a strong Release process does. CMDB in and of itself does not predict success, but linking changes to CIs in a CMDB does (which I would argue in turn justifies the CMDB - it's all about what you do with it, not just having it).
The performance measures seem primarily operational (a full list is not available in the summary). That is, "performance" is defined in terms of uptime, successful releases w/o rollbacks, etc - not in terms of business performance, IT porfolio alignment, etc.