Many organizations assume that tapping directly into operational technology data will reveal hidden insights that save huge amounts of money, but the reality of process data is far messier than they imagine. Measurements are full of limitations and artifacts that depend on specific instruments, installation conditions and physical constraints. A simple concept like “flow” hides complex behavior: Venturi flow meters only maintain their advertised accuracy across a limited operating range, their performance degrades badly when pipes are oversized for extreme conditions, and installation shortcuts such as placing them next to bends or valves introduce significant, unpredictable errors. Entire professions exist to understand these subtleties, and operations and engineering teams routinely correct and interpret imperfect readings based on long experience. Pulling raw values straight from the field without that context leads to confusion, mismatched totals and wasted time while data scientists struggle to rediscover what plant personnel already know.

The closer external systems get to process data, the more strain and risk they impose on the control environment. Controllers and field devices are built to run real-time processes first and share data only in the narrow time windows that remain; heavy polling or poorly planned data collection can interfere with communications to other control elements or operator HMIs, provoking serious operational backlash. On top of that, any connection into control networks must be made fail-safe and rigorously secured, because misuse or compromise can cause physical consequences and dangerous process upsets. A better approach invites data scientists into the plant to talk with superintendents, engineers and operators, understand how instruments are sized, where anomalies arise, how data is corrected, and where network capacity is already near its limits. Only by starting with questions and respecting these constraints can data initiatives avoid overloading systems, misreading noisy signals, and turning a data lake into a costly, misunderstood tar pit.

‍

Source: http://scadamag.infracritical.com/index.php/2025/11/04/the-truth-about-ot-data-and-what-it-costs/