Essay
A group can make sound technology decisions in every business and still be unable to answer basic questions about itself. The work is deciding which few things must be common.
A group I worked with operated in four sectors. Every one of its businesses had made defensible technology decisions for itself. Nothing had been bought foolishly, nothing had been neglected, and each managing director could give a coherent account of why their estate looked the way it did.
Collectively, the group could not answer basic questions about itself without a manual exercise.
That gap is the thing worth understanding, because it is not a technology failure. Every local decision was sound. The group had simply never decided which things had to be the same, and a set of individually correct decisions does not add up to an answerable organisation.
The work is choosing the layer
A transformation programme is an attempt to change how an organisation makes and executes decisions. Software is the instrument, not the subject. When it is framed the other way round, as a technology rollout with change management attached, it produces a modernised version of the same dysfunction at considerable cost.
In a group, that general principle becomes a specific and answerable question: which small number of things must be common for the group to function, and which differences are legitimate and should be left alone?
Most programmes never ask it, because there is a default answer that sounds like leadership: standardise everything. One platform, one process, one way of working. It is easy to announce, easy to fund, and it removes the need to make any further judgement.
Why uniformity is the easy answer and usually the wrong one
The difficulty with a multi-sector group is that the businesses are genuinely different. A healthcare operation and a construction contractor do not want the same system and should not be forced into one for the sake of tidiness. Their cost structures differ, their regulators differ, their operational tempo differs, and the things that constitute good performance differ.
Mandating uniformity across that does not produce a coherent group. It produces a system that fits none of the businesses well, a set of local workarounds that quietly restore the original variation, and a reporting layer that is now wrong in a more confident-looking way.
Standardisation is a tool, not a virtue. Used on the right layer it makes a group governable. Used everywhere it is just expensive conformity.
Where the line actually falls
The line I work to is this: something must be common where accountability crosses a boundary, and may legitimately vary where it does not.
If the group, the board, the auditor or the regulator is answerable for it, it has to mean the same thing in every business. If the accountability sits entirely inside one business, that business should be allowed to decide, and should be defended when it does.
Applied to a real group, four things sat above that line.
Financial definitions. Not the chart of accounts as a technical artefact, but what revenue means, when it is recognised, what sits in cost of delivery, and how intercompany work is treated. Without this the consolidated numbers are an opinion.
Governance. Who may commit the group to what, at which threshold, and through which forum. This is the layer that makes a decision traceable afterwards, which is the difference between a group and a holding structure.
Security posture. Risk does not respect the subsidiary boundary. A group is exposed at the level of its weakest business, so the standard has to be set centrally even where the implementation is local.
Reporting. Not the dashboards, which can differ, but the definitions underneath them, so that the same word means the same thing in four sectors.
Operational systems stayed local. The healthcare business kept what suited healthcare. The contractor kept what suited construction. That is a less impressive-sounding programme than a single-platform consolidation, and a considerably more successful one.
The part that is actually hard
Deciding the line is analytical work and can be done in weeks. Holding it is political work and never finishes.
Leaving things alone is harder than mandating uniformity, which is counterintuitive until you have tried it. A group executive who standardises everything is seen to be acting. A group executive who says that four businesses may keep four different maintenance systems has to explain that position repeatedly, to people who suspect it is indecision, and has to keep explaining it every time a vendor offers a consolidation discount.
It is worth the argument. Variation that has been examined and deliberately permitted is not the same condition as variation nobody ever looked at, even though the estate diagram looks identical. The first is a decision. The second is the absence of one, and it is the second that produces a group unable to answer questions about itself.
The sequencing trap
There is a related failure that shows up before any of this is settled. Organisations begin with the system that is easiest to buy rather than the process that is most broken, because the first has a vendor advocating for it and the second has nobody. Two years later the difficult thing is still difficult and the budget is gone.
The common layer is exactly the sort of work that has no vendor. Nobody sells financial definitions. Nobody runs a campaign for governance thresholds. If the programme is allowed to be driven by what is procurable, that layer will be skipped, and the group will end up with better software sitting on top of the same unanswerable question.
A test you can run this week
Pick a question the group ought to be able to answer about itself. Total committed spend with a supplier used by three businesses. Headcount by function rather than by entity. Open safety incidents across the group.
Then watch what happens. If someone can answer it from a system, the common layer exists. If the answer arrives two days later as a spreadsheet assembled by four people, it does not, and the gap is not in the reporting tool. It is in the definitions underneath, and no amount of analytics spend will close it, because the analytics will faithfully reproduce four different meanings of the same word.
That manual exercise is the most reliable diagnostic I know. It is also the one most often mistaken for a reporting problem.
How I take this on, and on what terms, is set out under advisory. What I hold directly rather than advise on is set out under leadership.