Manufacturers Can Find Their Data But Cannot Certify It

Cloudera released manufacturing findings from its Data Readiness Index 2026 on 8 September, and two numbers describe the state of industrial AI better than most surveys manage.

Eighty-two percent of manufacturing respondents report visibility into where their data resides. Fifty-eight percent say all or nearly all of that data is fully governed. Twenty percent name weak AI and analytics integration into operational workflows as the leading reason initiatives fail to deliver expected ROI. Cloudera attributes the difficulty to data and processes scattered across factories, supply chains, enterprise applications and edge environments. Morgan Bowling, Cloudera's director of global industry AI solutions for industrial and manufacturing, said success requires confidence in the quality, governance and availability of data across the business.

The 24-point spread between those first two figures is the whole story. Knowing where a tag lives is not the same as being able to say what it means, when it was last calibrated, whether the unit changed after a controller upgrade, and who is responsible when it drifts. A model can be trained on data nobody can certify. It cannot be put into a control loop or a quality decision on that basis, and that is precisely where most pilots stall — not in the modelling, but at the point where someone has to sign that the input is trustworthy.

The practical implication points away from moving more data to the cloud and toward contextualisation at the edge: naming, units, provenance and quality attached to the signal where it originates, rather than reconstructed downstream by someone reading a tag list. That work is unglamorous and historically hard to fund, which is the other thing a number like this is useful for.