Why Most Analytics Transformations Fail Before They Start.

Let’s skip the part where I cite a McKinsey statistic and you nod politely. You already know the failure rates are brutal. You’ve probably lived one. Maybe you’re living one right now, reading this on your phone, in a meeting about your organisation’s data transformation strategy, while a consultant explains what “data-driven” means.

The question worth asking isn’t why do analytics transformations fail. It’s why we keep being surprised when they do and why we keep starting the next one the same way.

The wrong question, asked confidently.

Here is how it usually starts. A leadership team returns from a meeting with investors, energised, aligned, newly fluent in the language of AI and data maturity and decides the organisation is going to transform. A strategy deck appears. A budget gets allocated. Someone gets a new title. And somewhere in the building, the people who actually know how the data works exchange a look.

That look contains an entire post-mortem.

Because the question driving all of this is almost always what capabilities do we need to build? The question that should be driving it is which decisions are we consistently getting wrong, and what would it actually take to get them right?

Those sound similar. They produce completely different outcomes. And most organisations, if they’re honest, have never properly answered the second one.

One leads to a technology roadmap. The other leads to a reckoning with data quality, with decision culture, with the gap between what leadership believes about the organisation and what the data would actually show if anyone trusted it enough to look. That reckoning is uncomfortable. A technology roadmap is easier to sell, easier to fund, and considerably easier to declare complete. So that’s what most organisations build.

The result is predictable: beautiful infrastructure, low adoption, and governance frameworks that get treated the way most people treat their gym membership in February, paid for with great intentions, rarely visited, and quietly forgotten until something goes wrong and everyone agrees to do better. Somewhere in the building there is a dashboard that took six months to build, that three people use, that nobody fully believes, and that appears in the board pack every quarter because removing it would require a conversation nobody wants to have.

AI is a brilliant tool. Used well.

This is not an argument against technology. AI is genuinely, meaningfully good at replacing low-risk, repetitive tasks — when the process behind them is well understood and the data feeding them is clean and trusted. In those conditions, it does exactly what it says on the tin. Less manual labour, fewer errors, faster throughput. Right tool, right job. The organisations using it that way are seeing real returns and good for them.

The harder conversation is about sequencing. The pressure to move fast on AI is real. It comes from boards, from competitors, from the specific anxiety of watching someone else’s press release and wondering if you’re already behind. That pressure has a way of compressing the preparation work that makes any of this sustainable. Foundations get assumed rather than built. Readiness gets declared rather than demonstrated.

And in that rush, workforce decisions get made that look entirely rational on the day they’re announced. The spreadsheet is clean. The narrative is compelling. The savings are real.

The cost shows up around month fourteen. Because what gets optimised away in those moments is rarely just a function. It’s almost always also a memory.

What leaves the building when people do.

Organisations are made of two things: the systems that store information and the people who understand it. The second category is significantly harder to replace than most restructuring plans acknowledge. Harder, and considerably more expensive, though that part tends to get discovered retrospectively.

The analysts, the coordinators, the person who has been reconciling that report manually every fifth of the month for four years, they carry something no platform migration captures and no onboarding document replaces. They know why the data looks the way it does. Which business rule was hardcoded during a merger and never revisited. Which metric leadership quotes in every board meeting and which one actually tells you what’s happening. They know the column marked “confirmed revenue” has meant three different things since 2019 depending on which team filled it in. They are, in practice, the connective tissue between the organisation’s messy operational reality and its official version of itself.

When that connective tissue goes, the new system inherits the complexity without the context. Things work, mostly. Until they don’t. And when they don’t, the people who would have known why are no longer around to say.

The people who remain are capable, often excellent. They are also now carrying more with less support. That combination doesn’t produce the step-change in performance a transformation promised. It produces something quieter and more corrosive. It produces a workforce doing its best while steadily running out of road, in an organisation that has reframed exhaustion as resilience and called it progress.

None of this shows up in the business case. The cost of reconstructing lost knowledge, of re-hiring capability that was rationalised away, of the eighteen months it takes to get someone new to the point the experienced person already was. That bill arrives later, attributed to other things, and paid for by someone who wasn’t in the room when the original decision was made.

Why the cycle doesn’t break.

The uncomfortable truth is not that organisations lack intelligence or ambition. Most leaders commissioning these transformations are thoughtful people, making real decisions under real pressure, with the information available to them at the time. That matters and it’s worth saying.

But the structural conditions around transformation create some patterns that are worth naming, because they repeat with remarkable consistency.

The timeline of a transformation and the tenure of the people leading it rarely align. Programmes that take three to five years to mature are regularly overseen by leaders whose careers move faster than that. Nobody designed it this way. It’s just how organisations work. What it means in practice is that the incentive tilts, consciously or not, toward visible momentum over quiet progress. Announcements over adjustments. Launches over landings. The bold move over the careful one.

Meanwhile, the people with the clearest view of what’s actually happening, the practitioners embedded in the work, the ones who could walk you through exactly where the gaps are, have often learned that raising difficult truths in a transformation context is a use of political capital most of them prefer to spend elsewhere. Is it cynicism? I don’t think so, it’s pattern recognition. So the feedback that would be most useful travels slowly, if it travels at all, and the programme proceeds on the basis of information that’s easy to share rather than information that’s true.

And because these failures tend to be slow and diffuse rather than sudden and visible, the connection between early decisions and late consequences rarely gets made clearly enough to change anything next time.

Which is, of course, how we end up here again.

What doing it right actually looks like.

The organisations that genuinely get this right are, almost without exception, boring about it. They don’t announce transformations. They fix one broken decision, measure it honestly, and do another. They treat data literacy as a years-long investment rather than a two-day workshop and a Teams channel. They hold on to the people who know where the bodies are buried because those people are, structurally and operationally, the most valuable thing in the building. They treat governance not as a compliance tax but as the architecture that makes everything else trustworthy.

None of that is glamorous. None of it makes a good keynote. You won’t see it in a vendor case study or a conference agenda. It also, quietly, works. And in five years, when the organisations that chose the louder route are on their third transformation and their second CDO, the boring ones will be making better decisions faster than anyone around them.

Turns out that was the point.

An honest invitation.

In my experience, the organisations that struggle aren’t struggling because the technology failed them. They’re struggling because the honest conversation that needed to happen before anything got procured never quite did.

That conversation isn’t technically complicated. It just requires the people who know what’s actually happening to feel it’s worth saying out loud.

So consider this an opening.

If you’ve watched the knowledge walk out the door, felt the cost that never appears in any ROI calculation, or sat in a meeting where you knew exactly what wasn’t being said, this is the space to say it.

Not the polished version. The real one.