What a Data Strategy Is, and the Decisions It Has to Make
A data strategy is a small set of choices about what data is for, what gets funded and what is left alone. The document matters less than whether those choices constrain anything.
A data strategy is the set of decisions an organisation makes about how data will serve what it is trying to achieve: which outcomes come first, which capabilities are paid for, who is responsible, in what order the work is done, and what is deliberately left out. Everything else in the document is explanation.
That definition is narrower than the one a strategy document can be written to. A paper that describes the value of data, lists the platforms the organisation intends to buy and closes with a set of principles has not yet decided anything. It can be approved without anyone giving something up, which is a reasonable test of whether it is a strategy at all.
This article sets out what a data strategy has to settle, how two published national strategies handle those questions, and how to tell a strategy that will change what people do from one that will sit on a shelf. It is the starting point for the rest of DII’s coverage of the subject.
What a data strategy is for
The UK National Audit Office reviewed the use of data across central government in 2019 and identified three conditions that have to be in place before data can be used well: “clear strategy and leadership; a coherent infrastructure for managing data; and broader enablers to safeguard and support the better use of data”1. Strategy comes first in that list because the other two depend on it. Infrastructure and safeguards cost money and attention, and somebody has to decide where both go.
The same report shows what the absence looks like. It found that government “does not treat data as a strategic asset”, and contrasted data with property, for which there were long-term funding plans and published strategies for government as a whole and for each department1. The comparison is useful outside government. An organisation that plans its buildings over ten years and its data one project at a time has decided, without saying so, which of the two it manages on purpose.
The purpose of a data strategy, then, is to make that management deliberate. The UK’s National Data Strategy states the aim in a phrase worth borrowing: a strategy should ensure that “we weigh the priorities and potential trade-offs of data in a deliberate and evidence-based way”2. Priorities and trade-offs are the substance. A document without either is a description.
What a data strategy is not
Three things can be presented as a data strategy and are something else.
A vision statement. A sentence about becoming data-driven, or treating data as an asset, states an attitude. It does not say which decisions will be made differently next year, or what will be given up to pay for them. A vision can sit at the front of a strategy. It cannot stand in for one.
A technology plan. A list of platforms, tools and migrations describes how some of the work will be done. It is silent on which business outcomes the work is for, so it cannot be used to choose between two projects that both look technically sound. Technology choices belong after the strategy and should be traceable to it.
A data management plan. Policies for quality, metadata, security and retention describe how data will be looked after. They are necessary, and they are a consequence of the strategy. They answer the question of how well data must be managed, which can only be judged once it is known what the data is needed for. The disciplines that set those rules are covered in what data governance is and what a working programme contains.
The distinction matters in practice because the data function can write each of these alone. A strategy cannot. It commits other people’s budgets and priorities, so it has to be agreed with the people who hold them.
The five decisions a data strategy makes
DII’s reading of the published strategies and of the audit evidence is that a data strategy has to settle five questions. The model is ours. The sources cited below support individual points, and none of them sets out these five as a list.
The five decisions a data strategy has to make
| Decision | The question it answers | What it constrains afterwards |
|---|---|---|
| Outcomes served | Which of the organisation’s objectives will data be asked to advance first? | Which use cases are in scope, and which requests can be refused |
| Capabilities funded | What will be built or improved, and what will it cost to run? | The investment case, the budget line and the skills to be hired or developed |
| Operating model | Who decides, who delivers, and what is done centrally? | Roles, accountability, team structure and the governance forums |
| Sequence | What has to come first, and what can wait? | The roadmap, its dependencies and the points at which it is reviewed |
| What is not done | Which plausible ambitions are being set aside for now? | The grounds on which later proposals are turned down |
1. Outcomes served
The first decision is which of the organisation’s objectives the strategy will serve, in order. The UK government’s data maturity framework describes the highest level of maturity in this area as one where the organisation “plans and prioritises data strategy as a core element of organisational strategy”, and the lowest as one where the two are linked “only where it is required for external reporting purposes”3. The gap between those descriptions is the gap between a strategy that begins from the business plan and one written beside it.
A short list is the point. Three or four outcomes that the executive team would recognise as its own priorities give the strategy something to be measured against. A list of fifteen leaves every later choice open. The method for getting from outcomes to the data work that serves them is the subject of how to align a data strategy to business outcomes.
2. Capabilities funded
The second decision is what will be paid for. The Federal Chief Data Officers Council’s playbook, in which serving chief data officers advise new ones, tells its readers to “link data specific resource allocations to the key business problems and objectives that are core to your data and analytics strategy”4. A capability with no budget against it is an aspiration, and the audit evidence on what follows is direct: the National Audit Office found one department where “the strategy has been approved but no additional funding has been agreed to implement it”1.
Funding decisions also have to cover running costs. A platform, a catalogue or a quality service needs people to operate it after the project that built it has closed, and a strategy that funds only the build has deferred the harder half of the decision.
3. Operating model
The third decision is who does what: which responsibilities sit with a central team, which with business units, and who has the authority to settle disagreements. Without it, the strategy has no owner for its commitments and no route for the decisions it will need later.
This is where a data strategy meets governance and the two should not be written separately. Who owns which data, and who may decide what about it, is set out in the governance arrangements; the strategy decides how much of that capacity the organisation is going to build, and where.
4. Sequence
The fourth decision is order. Some work makes other work possible, and some can be done at any time. A strategy that presents all of its initiatives as equally urgent has left the sequencing to whoever shouts loudest in the first budget round.
The published US Federal Data Strategy handles this by separating what is enduring from what is scheduled. It describes “a ten-year vision” and then commits, each year, to an action plan that will “identify priority actions for a given year and incrementally build on progress from year to year”5. Turning a set of decisions into that kind of sequenced plan is covered in from data strategy to roadmap.
5. What is not done
The fifth decision is the easiest to leave out, and DII regards it as the clearest sign that the other four are real. Every organisation has more good uses for data than it can pursue. A strategy that names the ones it is setting aside, and says why, gives managers a basis for saying no later without reopening the whole argument.
There is no external source for this point. It follows from the others: if outcomes are ranked, funding is finite and work is sequenced, then some things have been excluded, and it is better for the document to say which.
How two published strategies are built
Two national strategies are public and can be read in full. They are written for governments, and nothing here suggests an organisation should copy either. Their structure is instructive.
The UK National Data Strategy, published in 2020, rests on four pillars: data foundations, data skills, data availability and responsible data. From those it derives five missions, introduced with the sentence “from these pillars, we have identified five priority areas of action”2. The pillars describe the conditions that have to hold. The missions say where effort goes first. A reader can disagree with the choice of missions, which is evidence that a choice was made.
The US Federal Data Strategy has four components: a mission statement, ten principles, forty practices and an annual action plan. The action plan’s job is stated plainly. It is to “specify measurable activities to implement the practices and that are the priority for a given year, providing timeframes for implementation and identification of responsible parties”6. The principles are described as “timeless guidance”6; the action plan is where dates and names appear.
Both documents separate direction from commitment, and that separation is the transferable idea. Principles and pillars can be stable for years. Commitments have owners, dates and budgets, and are revisited on a cycle. A strategy that offers only the first kind of content cannot be checked, and one that offers only the second has no way to explain its choices.
Who has to own it
A data strategy commits resources that the data function does not control, so it needs an owner who does. The UK maturity framework’s description of an organisation at its fourth level of five says that “data strategy has sponsorship at a senior level”, and at the fifth that sponsorship “is visible and communicated”3.
Sponsorship is a weaker word than ownership, and the difference shows when priorities collide. The National Audit Office observed in 2019 that data projects “have sometimes been set aside when funding is under pressure”1. A sponsor can express regret when that happens. An owner with authority over the budget can decide whether it happens.
The strategy also has to be agreed by the people whose plans it changes. The US playbook’s fourth play describes a data strategy that can “influence the enterprise-wide business strategy”4, which is only possible if the people who write the business strategy took part. A newly appointed data leader inherits this problem on arrival, and the first 90 days as a data leader looks at how to handle it before a strategy is promised.
How to tell a strategy from a wish list
The following questions can be put to any draft. They are DII’s working tests and they are meant to be answered quickly.
- Can a reader name the three or four outcomes the strategy serves, and would the executive team recognise them as its own?
- Is there a figure for cost, including the cost of running what is built?
- Does each commitment have a named owner outside the document’s authors?
- Is it clear what happens first, and what depends on it?
- Does the document say what will not be done?
- Could a proposal be refused by pointing to it?
A draft that fails most of these is a statement of intent. That has its uses, particularly early on, but it should not be approved as a strategy, because approval will be taken as a decision that has not been made.
Where a strategy needs to be stated compactly before it is elaborated, the One-Page Data Strategy Template forces the five decisions onto a single page, which makes the missing ones obvious. The Data Strategy Alignment Guide supports the first of them, the tracing of outcomes to data work.
Where to go next
Each decision in this article is taken further elsewhere. Establishing a starting position, and its limits as a guide to action, is covered in what to assess in a data strategy maturity assessment and how to act on the result. What happens after approval, when a sound strategy stops moving, is the subject of why data strategies stall after approval.
A data strategy also sets the conditions for the reporting an organisation relies on. Whether figures can be trusted depends on decisions about definitions and ownership that a strategy either makes or leaves open, which is where what makes business intelligence trustworthy picks up.
Sources
- National Audit Office (UK). Challenges in using data across government (HC 2220). 21 June 2019.
- Department for Digital, Culture, Media and Sport (UK Government). National Data Strategy. 9 September 2020.
- Central Digital and Data Office (UK Government). Data Maturity Assessment for Government: Framework. 27 March 2023.
- Federal Chief Data Officers Council (United States Government). CDO Playbook: Advancing the Federal Data Strategy. 2021.
- Federal Data Strategy (United States Government). Federal Data Strategy 2020 Action Plan. 14 May 2020.
- Federal Data Strategy (United States Government). Federal Data Strategy: “Overview” and “Practices” pages. 2019.
This article is general professional information. It is not legal or regulatory advice.