Demystifying the language, rediscovering the fundamentals, and rebuilding trust.

Across aerospace, defence, and space, organisations are accelerating their adoption of AI to improve safety, reliability, and operational performance. As we rely more and more on AI automation, the quality of the data matters more than ever.

Data embraces every aspect of a business from decision making to overall operational efficiency and supporting relationships with customers and supply chain partners, so imagine the impact with bad data, and the potential magnification when AI is involved.

In this blog, we hear from Salesforce’s Graham Carr, who specialises in data, AI, and digital transformation within the aerospace and manufacturing sectors. He is also a leading member of the ADS Digital Transformation for Aerospace and Defence (DTAD) Special Interest Group.

AI has brought data quality back to the spotlight — but the fundamentals haven’t changed.

However fast technology moves, some things don’t change — and data quality is one of them.

Fifteen years ago, when I first entered the data space, Data Quality was an obscure corner of operations. Important, but largely out of mind. The topic felt dry, wrapped in specialist terms and frameworks that made it feel distant from everyday business life. Today, data-led decision-making is everywhere in an organisation. The fundamentals remain reassuringly familiar. Accuracy, completeness, consistency, and timeliness still define whether we can trust what we see.

In this article, I revisit those enduring principles with 10 core areas to consider. They show what ‘good data’ actually means and what impact bad data can have on AI processes.

1. Accuracy – Getting the Facts Right

Accuracy is the cornerstone of trust. It means that data correctly represents reality.

If an aircraft tail number is recorded incorrectly, every linked process — from maintenance tracking to billing — risks drifting off course.

Inaccurate data often starts small: a mistyped reading or a wrong code. Yet in AI systems, those small errors multiply, spreading flawed insights across multiple systems.

Good accuracy begins at the point of capture — through validation, feedback, and awareness.

Accurate data doesn’t just describe the world correctly — it allows confident decisions to be made without hesitation or second-guessing.

2. Completeness – Filling in the Gaps

Completeness means having all the essential information captured. For example, a maintenance log missing a date or technician name might look fine, but it’s incomplete — and therefore unreliable for analysis or audit.

Incomplete data is like a checklist with missing boxes — you can’t tell what’s done and what’s missing. In AI models, missing values skew results and reduce reliability.

Completeness isn’t about collecting everything; it’s about collecting what’s needed to tell the whole story. When every required piece is present, data becomes both trustworthy and actionable.

3. Consistency – Speaking the Same Language

Consistency ensures data is represented the same way wherever it appears.

An engine type recorded as ‘Trent 900’ in one system and ‘RR900-Trent’ in another may both be correct, but inconsistent. Consistency is the glue that lets systems connect and teams collaborate. It’s what allows AI models to recognise patterns and draw conclusions. By standardising formats, naming conventions, and definitions, organisations create a single language for data — one that allows confidence to flow across every team and tool.

4. Timeliness – Keeping Data Fresh

Even perfect data loses value when it’s out of date.

Timeliness ensures that data reflects the present, not the past. For example, flight delay information that’s accurate but published too late is of little use. Outdated data can quietly train AI systems on yesterday’s world, leading to poor forecasts or bad predictions.

Automating updates and flagging stale data keeps systems aligned with reality. Timely data sustains momentum and ensures decisions are based on what’s happening now — not what used to be true.

5. Validity – Following the Rules

Validity checks whether data fits expected rules or logic.

Valid data follows agreed formats and business rules so that systems can process it without exceptions or guesswork. For example, a date like ’31/02/2025′ or a negative temperature reading instantly signals invalid data.

In AI, validity acts like a first line of defence against corruption. Simple rule checks prevent false patterns and unreliable outcomes.

Validity doesn’t guarantee accuracy, but it does guarantee structural integrity — a foundation where meaning stays intact.

6. Uniqueness – One Record, One Reality

Uniqueness means each entity appears only once.

Duplicate customer records or supplier entries quietly distort reporting, billing, and AI training. Duplicates split histories, double-count events, and confuse automation. By assigning unique identifiers and automating duplicate detection, organisations protect the single version of truth.

When every record is unique, systems stay aligned, and decisions become simpler and clearer.

7. Integrity – Keeping the Links Intact

Integrity protects the relationships between data points. For example, a maintenance record linked to the wrong aircraft breaks integrity — even if each field is individually correct.

Integrity ensures related data stays correctly connected: parts to engines, customers to contracts, sensors to locations. When links stay intact, information retains meaning.

In AI systems, that integrity prevents false correlations and ensures reliable learning.

8. Reliability – Trusting the Source

Reliability reflects confidence in where the data came from and how it’s handled. Sensor data captured directly from an engine is generally more reliable than data rekeyed from a spreadsheet. Reliable data behaves consistently under consistent conditions. Strong processes, clear ownership, and automation help preserve reliability.

In AI, reliable data translates directly into credible outcomes — because trustworthy models depend on trustworthy inputs.

9. Lineage – Knowing Where It Came From

Lineage means being able to trace the data’s journey ie where it originated, how it was transformed, and how it’s being used. Without lineage, organisations operate on faith. Errors can’t be traced, and bias can’t be explained. In regulated industries like aerospace, unclear lineage can even breach compliance.

Lineage provides transparency — the ‘service history’ of data. In AI, it enables accountability and explainability. When you can trace data’s path, you can stand by its conclusions.

10. Relevance – Fit for Purpose

Relevance ensures that data actually serves the purpose it’s used for and ensures it’s meaningful. For example, perfectly accurate and complete data about retired engines might be technically sound but irrelevant for predicting future maintenance. Relevance connects quality with purpose, ensuring data is not only right, but right for now.

Relevance also depends on context: what’s useful for one decision might be noise for another.

Conclusion

As we embrace AI, these 10 ‘old truths’ have returned with new urgency. Bad data affects the whole organisation and everyone can play their part in ensuring that the insights they rely on are built on truth, not assumption.

Good data, like good engineering, lasts because it’s maintained with care. Accuracy without completeness is fragile. Consistency without timeliness is outdated. The real power of data comes when all these dimensions work together, forming an ecosystem of trust.

The fundamentals of data quality haven’t changed, but the stakes have.

Graham Carr – Salesforce Solution Engineering – Aerospace & Defence

In his role at Salesforce, Graham specialises in data, AI, and digital transformation within the aerospace and manufacturing sectors. He has held operational leadership roles at Rolls-Royce, National Grid, Veolia, and Meggitt, where he helped design next-generation service and maintenance models. He now helps organisations build trust in their data and apply AI responsibly — connecting decades of engineering experience and operational leadership with the realities of modern information management.

SUPPORT FOR ADS MEMBERS

ADS Digital Query Platform – Your Peer-to-Peer Advice Tool

Graham is an Executive Committee member of the Digital Transformation for Aerospace and Defence Group, an ADS Special Interest Group with a mission to actively inform, facilitate, and foster the widespread adoption of digital technologies throughout the entire lifecycle of products and services within the aerospace and defence sectors.

The group has launched its Digital Query Platform – a member-to-member advice service designed to help you navigate digital challenges and opportunities.

How it works:

  • Ask a Question – Submit your query via a short online form. Topics could include adopting new technologies, streamlining processes, or overcoming digital challenges.
  • Connect with Experts – Your question will be shared with contributing ADS members who have volunteered to provide practical advice and insights. View full list below.
  • Get Tailored Advice – Receive guidance from peers with relevant experience to help you find solutions.
  • Explore Opportunities – Advising members can also connect with you to discuss collaboration or business development.

Find out more about the platform, and submit your questions, here.