Data strategy

Most documents titled "data strategy" are technology plans wearing a strategy cover. They list platforms, capabilities and phases. They rarely say what the organisation has decided not to do, and almost never say who will be held to account if a named decision does not get better.

A strategy is a set of choices under constraint. If a document contains no choice that closes something off, it is an inventory.

What a data strategy is for

The audience is the executive committee, and the job is to make the next investment argument shorter than the last one.

Concretely, it exists to answer four questions in a form a board can hold: which decisions are we trying to make better, what are we deliberately not doing, what order does the work happen in, and who owns each outcome. Everything else, the architecture, the tooling, the delivery method, follows from those answers and can be changed without reopening them.

That is the test of whether a strategy is doing any work. If a vendor decision, a reorganisation or a budget cut can be resolved by pointing at the document, it is a strategy. If each of those events triggers a fresh round of debate from first principles, the document was an inventory.

Why most data strategies fail

The common failure is not ambition, it is the absence of exclusion.

A strategy that promises trusted data, self-service analytics, a modern platform, embedded governance and an AI capability has not prioritised anything. It has listed the field. Under budget pressure, which arrives eventually, there is nothing in the document that says which of those five survives, so the decision gets taken by whoever holds the budget at the time, on grounds the strategy never recorded.

The second failure is treating the estate as uniform. Governance applied evenly across every table spends the same effort on a regulatory submission as on a dataset nobody makes a decision with. That is expensive and it is also slow, because the work that actually carries consequence sits in a queue behind work that does not.

The third is measuring delivery rather than consequence. Dashboards shipped, datasets catalogued and models deployed are all real numbers, and none of them says whether a decision changed.

The four choices

Scope. Name the decisions. Board reporting, regulatory submission, pricing, credit, capital adequacy, channel mix. A decision has an owner, a cadence and a consequence for being wrong, which is what makes it a usable unit of scope. A capability does not.

Exclusion. Write down what is out, and for how long. This is the part that gets cut in review and it is the part that makes the rest enforceable. An exclusion with a review date is a decision; an exclusion left implicit is an argument waiting to happen.

Sequence. Order the work by what unblocks what, not by what is easiest to start. Most estates have one or two foundational gaps, usually ownership and lineage on a small population of critical fields, that quietly cap every other initiative until they are closed.

Ownership and stop rules. Each outcome gets a named owner in the business, and each initiative gets a condition under which it stops. Programmes without stop conditions do not stop, they get quietly defunded, which costs more and teaches the organisation nothing.

Strategy and the operating model

A strategy chooses. An operating model is what has to be standing for those choices to survive contact with the organisation, and the two are routinely confused.

The connection is direct: the decisions named in the scope choice become the top layer of the operating model, and every layer beneath, critical data elements, lineage and controls, stewardship, platform, assurance, exists to serve them. If a layer cannot name the decision above it, the strategy has not reached it yet.

That is why the operating models on this site are drawn as structures held at once rather than as phases. A strategy that hands over a sequence but no structure produces a programme that delivers and then decays.

Sequencing the work

Sequence is where a strategy becomes executable, and it is the point at which a framework earns its place. A framework is a movement through time with a gate between each stage, so it is the right shape for turning a chosen sequence into work that can be inspected while it is happening.

The practical pattern is the same in most regulated estates. Diagnose the evidence rather than the architecture. Frame the gaps in terms a board can act on. Decide the order and, explicitly, what is being deferred. Embed ownership so the work survives the end of the funding.

None of that requires a new methodology. It requires that the sequencing decision is recorded somewhere an executive can point to six months later.

Knowing whether it worked

The measures that matter are decisions changed, time from question to defensible answer, and the cost of proving a number when someone challenges it.

Usage statistics are worth reporting and are never the headline, because a heavily used dashboard that changed nothing is a cost that looks like a success. The same applies to catalogued datasets and to models in production.

A strategy that cannot say, twelve months in, which decisions are now made differently has not failed to deliver. It has failed to choose.