Why Data Risk Often Emerges Late in Transformation

In many transformation programmes, the most significant data challenges don’t appear at the start. They emerge during testing, reconciliation, or cutover rehearsal, precisely when the programme is least able to absorb them.

This isn’t unusual. It’s a pattern TDMC has seen across industries and system landscapes. And it’s supported by industry research:

IBM reports that 70% of organisations experience data errors during cloud migration (IBM, State of Cloud Migration, 2022).

McKinsey highlights that 60% of cloud migrations struggle due to legacy integration issues (McKinsey, Cloud Migration Insights, 2023).

These issues rarely originate in the new system.

They originate in the legacy estate.

1. Early stages focus on the target, not the truth in the data

Transformation programmes naturally prioritise:

• target architecture
• new capabilities
• vendor alignment
• delivery milestones

Legacy data is often assumed to be “manageable”, something that will be addressed once the new system is ready. But legacy data carries history, and that history isn’t always visible in documentation.

As TDMC co-founder Arun Atri notes:

“Hidden rules and legacy behaviours often only reveal themselves when the data starts to move.”

This understanding often comes later than teams expect.

2. Documentation doesn’t capture behavioural drift

Systems evolve.
Teams adapt.
Workarounds accumulate.

Documentation rarely reflects:

• how fields are used today
• how values have changed over time
• how different systems interpret the same data
• how operational teams have adapted processes
• how historical fixes became embedded behaviours

These realities only emerge when teams interrogate the data directly, often during testing or reconciliation.

3. Hidden rules surface when pipelines are built

Across TDMC’s migrations, we consistently see:

• values behaving differently across systems
• dependencies that were never documented
• business logic embedded in data rather than code
• inconsistent master records
• divergent behaviours across markets or business units

These issues become visible only when pipelines start moving real data.

By that point, the programme is deep into delivery.

4. Cutover pressure amplifies every issue

As Arun explains:

“Cutover is the moment where every decision becomes real. Systems pause, data moves, and the business feels the impact of any unresolved issues.”

When data issues surface at this stage, they affect:

• go‑live readiness
• customer experience
• regulatory compliance
• revenue assurance
• operational continuity

This is why organisations often see the real data risk late, not because they weren’t looking, but because certain behaviours and dependencies only reveal themselves under real movement and real pressure.

5. How TDMC shifts discovery earlier in the lifecycle

TDMC’s approach is designed to uncover these behaviours long before cutover, by examining how the data operates across the legacy estate rather than waiting for issues to emerge during testing or rehearsal.

We do this by concentrating on four areas:

• Interrogating the data directly rather than relying solely on documentation or system behaviour. This exposes how entities actually behave across systems, not just how they were intended to behave.
• Comparing the same entities across multiple systems to reveal inconsistencies such as divergent field usage, behavioural drift, and inconsistent master records. These differences often explain why data behaves unpredictably during testing or cutover.
• Identifying business logic embedded in the data itself, including historical workarounds, value‑based rules, and operational patterns that may not appear in process documentation or code.
• Highlighting where legacy data structure may challenge target‑side concepts, by examining how records are organised, related, or distributed across systems. This helps identify areas where the business may need to make decisions or adjustments long before detailed design or mapping begins.

As Arun explains:

“Early visibility comes from understanding how the data behaves today, not how it was documented years ago.”

This shift in discovery reduces late‑stage surprises and supports more predictable delivery.