As adoption of industrial AI accelerates, many data leaders are facing a common challenge. Detecting and prioritising OT data quality issues is an essential first step towards AI readiness. But finding the problems is only half the job. Data issues can’t always be fixed at the source (for example sometimes ownership sits with another team, or the data needing remediation is historical) but the data still needs to be cleaned up before it’s any use for an AI model or a report.
DataWise is already the OT data trust layer for some of the world’s biggest industrials, continuously detecting anomalies across every channel and scoring them using our proprietary metric, the Data Quality Index (DQI).
Version 4.5 is where detection turns into action, extending DataWise in three directions:
- It proactively cleans unnecessary noise from your data, through Data Remediation.
- Albert, DataWise’s agentic data quality expert prioritizes issues and performs root cause analysis, all through plain conversation.
- It empowers different teams to detect and prioritize data quality issues according to what matters most to them, through DQI Configuration.
Data Remediation: turn detected issues into clean, usable data
Data Remediation solves two problems at once. The first is time: data teams spend endless hours cleaning data by hand, and because they often sit a step removed from operations, much of that effort goes on simply working out which anomalies are real issues worth reporting, and which are just noise from a sensor fault or an IT artefact that needs clearing.
Remediation lets them clear that noise in one pass – you tell it which kinds of issues to strip out, and it cleans them while leaving the real signal untouched. The second is AI readiness: industrial AI models are only as good as the history they learn from, and Remediation is what turns a messy historical record into clean data teams can confidently build on. The result is a remediated copy of your dataset you can feed to your AI models with confidence. Because you decide what to remediate and what to leave in place, one clean data source has multiple uses.
Figure 1: A remediation rule defines which issues to act on (E.G. by engine and severity) and what to do with each flagged value. Here, Synthesize regenerates flagged points from the surrounding signal pattern.
- AI and analytics – fill No Data and Flat Line gaps to produce a continuous series, so missing data doesn’t break a model in training or inference. For example: a vibration sensor on a critical pump drops offline for several hours overnight, leaving a No Data gap, and a temperature channel flatlines when its transmitter freezes. Fed straight into a predictive-maintenance model, those gaps would corrupt the training window and distort its forecasts. A rule using the Synthesize method regenerates each missing point from the surrounding signal pattern, producing a continuous series the model can learn from, and flags every filled value as synthetic, so the data lead knows exactly which readings were reconstructed.
- Reporting and compliance – remediate only high-severity issues, such as clearly erroneous readings, so a bad value doesn’t skew a regulatory or monthly report while smaller variations stay intact. For example: a flow sensor logs a handful of clearly erroneous spikes during a calibration glitch. Left in place, they would inflate the totals in a monthly compliance report. A rule targeting only high severity Out of Range and Bad Values events corrects those points in the reporting copy, leaves genuine process variation untouched, and flags every change, so an auditor can see exactly what was adjusted.
One principle sits underneath all of it: every synthesized value is flagged as synthetic, never passed off as an original reading. You always know which points were remediated and which came straight from the source, so the cleaned dataset stays fully auditable.
Figure 2: Rules are reusable and purpose-built. Each targets a different combination of engines, so the same source data can be prepared for a report, for a model, or to clear IT noise — without touching the historian.
Albert: know what to act on, and why
Remediation gives you a way to fix issues, but on a large estate, thousands of data quality events can surface at once, and busy teams struggle to know where to focus. Albert is built for to help manage this increasingly common scenario as higher volumes of OT data are captured and analyzed.
Albert is DataWise’s AI data quality expert: A conversational assistant that replaces moving between screens, building queries, and reading dashboards. You ask a question in plain, operational language — “What are my most urgent issues right now?” or “Can I trust the data on line 3 today?” and Albert works through your data quality events and DQI to answer. In practice, it does three things well:
- Triage – ask what needs attention first, and Albert surfaces the top issues, ranked by severity, duration, how many channels are affected, and the equipment involved. It leads with a short summary, then the detail.
- Trust assessment – ask whether a given area, line, or channel is reliable right now, and Albert tells you whether the data is missing, drifting, out of range, or inconsistent, which engines are driving the loss of trust, and whether the problem is new or recurring.
- Interpretation – Albert explains what an engine, event, or DQI score means in plain operational terms, and groups recurring patterns across your worst-performing channels, so a flatline cluster on a set of pumps reads as one pattern rather than fifty separate alerts.
Figure 3: Ask Albert what needs attention, and it ranks issues by severity, duration, and channels affected, groups related events into clusters, and leads with the takeaway.
What makes those answers usable for decisions is the discipline behind them. Albert draws only on DataWise’s own data, not guesswork; when it isn’t sure, it says so rather than inventing a tag or a number; and you can always ask it to show its reasoning. An assistant that tells you what it doesn’t know is one you can actually rely on.
The result is that the slow, expert-dependent part of data quality management that works out what is happening and what deserves attention first, becomes something any engineer can do in a quick conversation, whatever their experience level.
DQI Configuration: measure data quality on what matters most to each team
Aperio’s proprietary Data Quality Index (DQI) has always measured data quality over time. But one unified score isn’t necessarily useful to every team. The data quality that matters to an executive tracking overall trust across the enterprise isn’t the same as what a process engineer needs before acting on a signal, or what a data lead needs before feeding time-series data into a model.
DQI Configuration lets you define more than one DQI dimension, each watching the engines and severity ranges that matter for a specific role. For example:
- An executive’s organization-wide DQI might track every engine at high severity, for a clean view of overall data trust.
- A process engineer’s DQI might track only the issues that would change a decision on the floor, at the severities worth acting on.
- A data lead preparing signals for AI might track No Data, Flat Line, and Sample Rate, where gaps and irregular sampling matter most.
- A data engineering team mid-migration might track a DQI built around Consistency, No Data, and Sample Rate, as these are the signals that surface silent dropouts, sampling imbalance, and delivery gaps as data moves from the historian to the cloud.
The out-of-the-box DQI keeps watching every engine at every severity as your baseline. From there you can add one or more additional DQIs, add them to your dashboard, share it with your team, and create as many as you need. This ensures data quality becomes measurable on the terms that each team recognizes.
Why DataWise 4.5 matters
DataWise has evolved from detecting and scoring data quality to acting on it, with transparency at every step. You can remediate issues without touching the source, lean on Albert to know what to act on first, and measure trust the way each team defines it.
This evolution represents the next generation data trust layer for industrials, making OT data decision-ready before it ever reaches a model, a report, or a decision.
Want to see how Aperio DataWise 4.5 transforms the way you work with your own OT data? Book a demo with us to see DataWise in action.