He makes the point that in fact that there is, in fact, a dearth of reliable data on things that matter and an embarrassment of data on things that don’t* (see: the Twitter firehose).
I think this thought. My pet view is that analytical capabilities are becoming more and more commoditized and that real ‘value added’ in the future will depend, to a greater and greater degree, on building data sets.
*Obviously, by “matter” I mean things of non-promotional economic value. The Twitter data could “matter” a helluva lot to advertisers.