Data Mirages: Are You Making Long-Term Decisions on Data You Can't Fully Trust?
By Shalini Sood, Senior Consultant (UK)
Right now, somewhere in a healthcare system, a board is approving a multi-million pound investment based on a dashboard. A transformation programme is being designed around demand projections. An AI model is being trained on a dataset. A workforce strategy is being shaped by activity data.
All of it looks defensible because the numbers reconcile, trends are clear and the reports tell a coherent story. But beneath that surface, the data is not always as solid as it appears. This isn't a reporting problem, it's a strategic risk problem.
When data fails, you can’t see it from looking at the dashboard. It fails at the outcomes of the decisions the dashboard informed, sometimes years after the fact, when the consequences have already compounded.
Why the surface looks clean when it isn't
Most data systems are built for reporting, not interrogation. Aggregation smooths out inconsistencies. Dashboards standardise what gets presented. Governance processes check submissions, not the assumptions behind them.
So what you end up with is something that looks consistent, but the information isn't necessarily all it seems.
Even mature systems acknowledge this, despite validation there is always some uncertainty in the true position of the data. Which creates a subtle but important way of understanding what you’re seeing, directionally useful data, but not decision-grade reliable. However, over time, organisations start relying on it anyway, not because they fully trust it, but because it's the best version available.
That is the mirage.
What good data due diligence actually looks like
We tend to treat due diligence as a validation step, a quick check to confirm that data is good enough. In reality the goal of due diligence isn't to confirm the data, it should be to challenge it.
Good data due diligence isn't about reviewing outputs. It's about interrogating what sits beneath them. It means asking uncomfortable questions and not stopping at the first reasonable answer.
Four questions important questions to ask:
where has this data come from and what has happened to it along the way?
are we all talking about the same thing when we use the same metric?
what is missing and are we aware of it?
does this data actually reflect how things work on the ground?
This is the work we do at Tektology before transformation programmes commit to direction. Not a validation pass on the data that already exists, but a deeper interrogation of what it can and cannot support. Because the cost of skipping that step rarely shows up quickly. It shows up when a programme is already too far in too course-correct.
The compounding cost of getting it wrong
The impact of poor data doesn't show up overnight or in year one, it builds. At the front line, it can affect patient safety due to missing details, inconsistent records, incomplete histories. But at a strategic level, the effects are harder to spot and often more damaging and expensive because of the time it takes for the problems to show up clearly.
Investment decisions get made on partial truths. Capital is committed against demand projections that don't quite hold up. Service redesigns are built on activity profiles that misrepresent how things actually work. Transformation programmes struggle for years because the foundation underneath them was never as solid as the business case suggested. Teams end up spending more time fixing data than using it.
As organisations lean further into AI and advanced analytics, the stakes or mistakes can increase. Because these systems don't fix bad data. They scale it, turning yesterday's quiet uncertainty into tomorrow's automated decisions, made faster and at greater volume than anyone can meaningfully review.
Why this matters now
The danger in healthcare data isn't obviously bad data. It's data that looks good enough to trust and gets trusted with decisions that shape years of investment, service design and patient outcomes.
In a world where those decisions are getting faster, more automated and more consequential, the cost of the mirage compounds. The organisations that will navigate this well are the ones willing to look harder at what their data is actually telling them and to bring in the right people whose job is to ask the questions their own systems are not designed to answer.
Because across complex healthcare environments, one thing shows up again and again: data works well in silos, but struggles at a system level. Clinical platforms capture detailed patient information. Operational systems track flow and activity. Financial systems manage cost and performance. Each does what it was built to do.
Unsurprisingly when you try to bring it all together, the cracks appear. Definitions don't align, granularity differs and reporting practices vary. Suddenly you are looking at multiple versions of the same reality, where what is accurate in one context becomes misleading in another. And even when data appears complete, it is rarely aligned across systems. This isn't unusual. It's structural.
Fragmentation runs deeper than technology
This level of fragmentation often gets framed as a technology problem. If we integrate the systems better, standardise the platforms, introduce new tools and it will improve. To some extent, that's true. But the issue usually runs deeper.
It comes from how organisations are structured, how KPIs are set, how processes evolve over time and from legacy decisions that were right at the time but no longer fit.
Fragmentation is the natural outcome of growth and complexity. Which is why it persists even in highly digitised environments. And why data due diligence so often surfaces issues that technology alone cannot solve. The temptation is to work around it rather than do a deep dive into solving and truly reconciling your data.
From data quality to data confidence
The bigger question worth asking isn't whether you have good data. It's whether you can trust this data enough to act on it — and to keep acting on it as decisions compound over time.
Data quality isn't just about accuracy or completeness. It's about confidence. Confidence that the data holds up across contexts. Confidence that it means the same thing to different people. Confidence that it can support real decisions, not just reporting.
Getting there requires a shift in approach. Making data lineage visible, not assumed. Aligning definitions across systems, not within silos. Strengthening governance where data is created, not just where it's reported. Treating due diligence as ongoing, not a one-off exercise. And ultimately, being willing to confront fragmentation rather than work around it.
Because in the end, the mirage doesn't disappear by looking away from it. It disappears when you decide to look harder and that is where the most consequential transformation decisions should begin.
Shalini Sood is a Senior Consultant at Tektology, based in London. With over six years of experience across healthcare consulting at EY, Charkos Global and now Tektology, she has worked across operating model design, workforce planning, digital strategy and performance optimisation — supporting clients in the UK, Australia, Canada and beyond. Her work focuses on the practical detail underneath strategic decisions: how data, processes and operational design support transformation.