Decision intelligence: from data-driven to decision-centric
Gartner defines decision intelligence as "a practical discipline that advances decision making through an explicit understanding and engineering of how decisions are made." In February 2026, they published their inaugural Magic Quadrant for Decision Intelligence Platforms, legitimising what practitioners have known for years: organisations do not struggle because they lack data. They struggle because they lack a structural model that connects data to the decisions that shape outcomes. This guide explores what decision intelligence means, where the current platforms fall short, and why metric trees are the missing layer between data and action.
9 min read
What is decision intelligence?
Decision intelligence is not a product category that appeared from nowhere. It is the convergence of several disciplines that have been developing independently for decades: decision science, causal inference, operations research, behavioural economics, and systems thinking. What changed is that Gartner, in its February 2026 Magic Quadrant, gave the convergence a name and a market definition. The causal AI market alone was valued at $336 million in 2025 and is projected to reach $1.1 billion by 2032. The investment signals are clear: organisations are moving beyond descriptive analytics toward systems that model how decisions produce outcomes.
Gartner's definition
Decision intelligence is "a practical discipline that advances decision making through an explicit understanding and engineering of how decisions are made, and the evaluation of outcomes based on feedback." It shifts the unit of analysis from the data point to the decision.
The distinction matters more than it first appears. For twenty years, the analytics industry has been organised around a single premise: give people better data and they will make better decisions. This premise built a multi-billion-pound industry of data warehouses, dashboards, and business intelligence tools. It gave every manager in every company a login to a reporting platform. And it failed, not because the data was wrong, but because the premise was incomplete.
Better data is necessary but not sufficient for better decisions. A marketing director staring at a dashboard with forty-seven metrics does not lack data. She lacks a model that tells her which of those metrics are causally connected to the outcome she is trying to achieve, which levers she can pull, and what the likely second-order effects of pulling them will be. Decision intelligence addresses this gap by making the decision itself the object of study, rather than treating it as something that happens automatically once the data is good enough.
The shift from data-driven to decision-centric
The phrase "data-driven culture" has been the north star of digital transformation for over a decade. It assumes a linear relationship: more data leads to more insight, which leads to better decisions. But the relationship is not linear. It is mediated by how people interpret data, how organisations structure choices, and whether anyone has mapped the causal chain between the numbers on the screen and the outcomes in the real world.
The decision-centric shift inverts the model. Instead of starting with data and hoping decisions improve, it starts with the decision and works backward to determine what data, models, and structures are needed to improve it. This is not a semantic difference. It changes what gets built, how teams are organised, and what success looks like.
| Data-driven approach | Decision-centric approach | |
|---|---|---|
| Starting point | Collect and organise data | Identify the decisions that matter most |
| Unit of analysis | The metric or data point | The decision and its outcome |
| Success measure | Dashboard adoption, query volume | Decision quality, outcome improvement |
| Key artefact | The dashboard or report | The decision model or causal map |
| Failure mode | Data-rich, insight-poor | Model-rich, adoption-poor |
| Who benefits first | Analysts and data teams | The people who make decisions |
Notice the last row. Data-driven initiatives tend to serve the people closest to the data: analysts, engineers, data scientists. Decision-centric initiatives, when they work, serve the people who actually make the choices that determine outcomes: operators, managers, and executives. This difference in who benefits first explains why so many data-driven transformations stall. The people with the most organisational power, the decision makers, often see the least immediate value from data infrastructure investments. Decision intelligence promises to reverse that by making the decision maker the primary beneficiary.
The challenge, as we will see, is that most decision intelligence platforms have not yet delivered on this promise. They have built sophisticated causal modelling and simulation capabilities aimed at data scientists, not the managers and operators who need them most.
The five components of decision intelligence
Decision intelligence is not a single technology. It is a discipline composed of five distinct capabilities, each addressing a different part of how organisations make and evaluate decisions. Understanding these components separately is important because most organisations already have some of them in nascent form, and the gap between where they are and where they need to be varies by component.
Decision modelling
Mapping the structure of a decision: what is being decided, who decides, what inputs are considered, what options exist, and what criteria determine the choice. Decision modelling makes implicit reasoning explicit. Most important decisions in organisations happen through an unstructured blend of intuition, politics, and pattern matching. Decision modelling does not eliminate intuition. It provides a scaffold that ensures intuition is applied to the right question with the right information.
Causal reasoning
Understanding not just what correlates with an outcome, but what causes it. Traditional analytics answer "what happened?" Causal reasoning answers "what would happen if we did X?" This is the capability that separates decision intelligence from business intelligence. A dashboard can show that conversion rate dropped after a pricing change, but it cannot distinguish correlation from causation. Causal models can, and that distinction is the difference between insight and actionable intelligence.
Simulation and scenario planning
Testing decisions before making them. Given a causal model, simulation allows decision makers to explore "what if" scenarios: what happens to revenue if we increase prices by ten per cent? What happens to churn if we reduce the support team? Simulation converts abstract models into concrete predictions that teams can evaluate, debate, and compare. It moves decisions from "we think this will work" to "our model predicts these outcomes with these confidence intervals."
Orchestration
Routing the right decision to the right person with the right context at the right time. Not every decision needs the same process. High-frequency, low-stakes decisions can be automated entirely. High-stakes, infrequent decisions need human judgement supported by models. Orchestration defines which decisions follow which path, ensures that decision makers have the data and models they need, and manages the handoff between automated and human decision-making.
Monitoring and feedback
Tracking the outcomes of decisions and feeding those outcomes back into the models that informed them. Without this loop, decision models become stale. With it, the organisation learns from every decision it makes. Monitoring also catches model drift: the gradual degradation of a causal model as the world changes and the assumptions embedded in the model become outdated. This is the component that transforms decision intelligence from a one-off modelling exercise into a continuous learning system.
These five components form a cycle, not a sequence. Decision modelling reveals which causal relationships matter. Causal reasoning quantifies those relationships. Simulation tests interventions. Orchestration ensures decisions reach the right people. Monitoring captures outcomes and updates the models. Each pass through the cycle improves the next. The organisations that benefit most from decision intelligence are not the ones that implement one component perfectly. They are the ones that connect all five into a continuous loop.
Where metric trees fit in
A metric tree is, at its core, a causal model. It decomposes a high-level outcome into the operational drivers that produce it, linked by relationships that are mathematical, causal, or both. Revenue decomposes into customer count and average revenue per customer. Customer count decomposes into new acquisitions and retention. Retention decomposes into product engagement, support quality, and perceived value. Each decomposition is a causal assertion: this metric moves because these metrics move.
This makes the metric tree a natural foundation for decision intelligence. It provides the structural model that connects decisions to outcomes. When a product manager decides to invest in onboarding improvements, the metric tree shows the causal chain: better onboarding should improve activation rate, which should improve retention, which should improve lifetime value, which should improve revenue. The tree does not just record what happened. It predicts what should happen if the decision is sound.
Each node in this tree is a decision point. Marketing Spend is a decision. Onboarding Completion reflects a product investment decision. Churn Rate is influenced by dozens of decisions across support, product, and pricing. The tree makes the decision architecture of the business visible. It answers questions that decision intelligence platforms are built to address: which lever has the highest impact on the outcome? Where is the system underperforming? If we change this input, what happens downstream?
Critically, the metric tree does this in a form that non-technical people can understand. You do not need to read a directed acyclic graph. You do not need to interpret a Bayesian network diagram. You navigate a tree. The simplicity of the structure is not a limitation. It is what makes the causal model accessible to the people who actually make decisions.
The connection
Decision intelligence needs a causal model that connects decisions to outcomes. A metric tree is exactly that model, expressed in a form that every person in the organisation can read, navigate, and act on. It is the bridge between the theoretical framework of decision intelligence and the practical reality of how organisations operate.
The gap in decision intelligence platforms
The decision intelligence platforms recognised in the 2026 Magic Quadrant are technically impressive. They can build causal models from observational data. They can run Monte Carlo simulations across thousands of scenarios. They can identify optimal decision policies using reinforcement learning. These capabilities are real and valuable.
But they share a common limitation: they are built for data scientists, not for the people who make the decisions. The typical workflow requires a specialist to define the causal graph, prepare the data, run the simulations, and interpret the results. The output is then presented to a decision maker in a meeting, often weeks after the question was first asked. By the time the analysis arrives, the decision has already been made on gut instinct, or the context has changed enough that the analysis is no longer relevant.
- 1
The accessibility gap
Decision intelligence platforms require technical fluency that most managers do not have. Building a causal model in these tools means understanding concepts like confounders, instrumental variables, and counterfactual reasoning. These are important ideas, but requiring every decision maker to master them is like requiring every driver to understand engine thermodynamics. The result is that the most sophisticated decision models sit unused because the people who need them cannot operate the tools that produce them.
- 2
The latency gap
Most business decisions are made in meetings, in Slack threads, in the ten minutes between calls. They are not made by submitting a request to a data science team and waiting for a simulation to run. Decision intelligence platforms operate on analytical timescales, not operational timescales. The models they produce are valuable for strategic planning but inaccessible for the hundreds of tactical decisions that compound into organisational performance.
- 3
The adoption gap
Technology adoption follows a predictable pattern: tools succeed when they fit into existing workflows rather than requiring new ones. Decision intelligence platforms typically require a fundamental change in how people work. They ask decision makers to formalise their reasoning, define variables, and think in probabilities. This is valuable practice, but it introduces friction that most operational teams will not tolerate. The tools end up as specialist instruments used by a small number of trained practitioners, not as organisational infrastructure that shapes how everyone decides.
- 4
The behaviour gap
The most overlooked limitation is that decision intelligence platforms model decisions as rational processes. They assume that if you give people the right information in the right structure, they will make better choices. Behavioural science tells a different story. People anchor on the first number they see. They overweight recent events. They avoid decisions that might produce regret, even when the expected value is positive. No causal model, however sophisticated, changes these patterns. Improving decision quality requires changing decision behaviour, and that requires a different kind of intervention.
These gaps do not mean decision intelligence platforms are without value. They are powerful tools for specific use cases: supply chain optimisation, fraud detection, pricing strategy, and other domains where specialist teams make high-value, repeatable decisions. But the broader promise of decision intelligence, improving how the entire organisation decides, requires something the current platforms do not provide. It requires a layer that sits between the sophisticated models and the people who need them, translating causal intelligence into a form that shapes daily behaviour.
Decision intelligence needs behaviour change
The insight that connects decision intelligence to metric trees is this: improving decisions at scale is not a modelling problem. It is a behaviour change problem. The models matter, but only insofar as they change how people actually behave when they face a choice. A causal model that sits in a data science notebook changes nothing. A causal model that is embedded in the structure people navigate every day, that shows them which metrics are connected to which outcomes, that alerts them when something upstream has changed, changes everything.
This is where the metric tree becomes the operational layer of decision intelligence. The tree takes the causal relationships that decision intelligence platforms model and embeds them in a structure that every person in the organisation encounters in their daily work. It does not require anyone to learn causal inference. It does not require anyone to run a simulation. It simply makes the causal structure of the business visible, navigable, and actionable.
Visibility replaces modelling
Instead of asking decision makers to build causal models, the metric tree presents the causal structure as a navigable hierarchy. A manager can see that their metric connects to the metrics above and below it. They do not need to understand directed acyclic graphs. They need to understand their branch of the tree, and the tree makes that understanding natural.
Alerts replace simulations
Instead of running what-if scenarios before every decision, the metric tree monitors the causal chain in real time and alerts owners when something changes. A drop in activation rate triggers attention to the downstream metrics it affects: retention, lifetime value, revenue. The tree simulates nothing, but it surfaces the same causal connections that a simulation would reveal.
Ownership replaces orchestration
Instead of routing decisions through automated workflows, the metric tree assigns ownership to every node. Each person knows which metrics they are accountable for, who depends on them, and whose metrics they depend on. Decisions about where to invest attention and effort are made continuously by the people closest to the work, guided by the structure of the tree.
Feedback replaces retrospective analysis
Instead of evaluating decision outcomes in quarterly reviews, the metric tree provides continuous feedback through leading indicators at the leaves and lagging indicators at the root. Decision makers can see the effects of their choices propagating through the tree in near real time, enabling rapid course correction without waiting for a formal review cycle.
The metric tree does not replace decision intelligence platforms. It complements them. The platforms provide the deep causal analysis needed for high-stakes strategic decisions. The tree provides the always-on causal structure that shapes the thousands of daily operational decisions that collectively determine whether the strategy succeeds. Together, they form a complete decision intelligence architecture: sophisticated models for the few decisions that justify them, and embedded causal structure for everything else.
The real power of this combination is that it addresses both sides of the decision quality equation. Decision intelligence platforms address the information side: giving people better models, better data, better predictions. The metric tree addresses the behaviour side: making the right information visible, creating feedback loops, establishing ownership, and embedding causal thinking into the daily rhythm of the organisation. Improving decisions at scale requires both. Better information without behaviour change produces unused models. Behaviour change without better information produces well-intentioned but poorly informed choices.
“Decision intelligence fails when it treats decisions as a modelling problem. It succeeds when it treats decisions as a behaviour problem. The metric tree is the behavioural layer that makes causal thinking operational.”
Getting started with decision-centric metrics
You do not need a decision intelligence platform to start thinking about decisions differently. You need a metric tree and a willingness to ask a different question. Instead of "what should we measure?" ask "what decisions matter most, and what would we need to see to make them well?" This reframing shifts metric design from a reporting exercise to a decision support exercise, and it produces a fundamentally different kind of tree.
- 1
Inventory your highest-leverage decisions
Every organisation has a small number of decisions that disproportionately determine outcomes. For a SaaS company, these might include pricing changes, feature investment prioritisation, and hiring allocation. For a retailer, they might include assortment planning, promotional strategy, and store staffing. List the ten decisions that, if made better, would produce the biggest improvement in your key outcome. This inventory becomes the design brief for your metric tree.
- 2
Map each decision to the metrics it affects
For each high-leverage decision, trace the causal chain from the decision to the metrics it influences. A pricing decision affects conversion rate, average revenue per user, and churn. A hiring decision affects throughput, quality, and cost. These causal chains become branches of your metric tree. The tree is no longer organised around what is convenient to measure. It is organised around what matters for the decisions you actually face.
- 3
Identify the metrics you are missing
When you map decisions to metrics, you will almost certainly discover that the metrics you currently track do not cover all the causal links that matter. You might find that you have no metric for time to value, even though it sits on the critical path between onboarding investment and retention. You might find that you track conversion rate but not the quality of the leads being converted. These gaps are where decision quality is leaking. Closing them is often more valuable than improving the metrics you already have.
- 4
Assign ownership based on decision authority
The owner of a metric should be the person who makes the decisions that most directly affect it. This sounds obvious, but in practice metric ownership often follows organisational hierarchy rather than decision authority. A VP may own a metric that is actually shaped by decisions made three levels below them. Aligning ownership with decision authority ensures that the person watching the metric is the person who can do something about it.
- 5
Review decisions, not just metrics
In your regular metrics review meetings, add a decision lens. When a metric moves, ask not just "why did this change?" but "what decision led to this change, and what did we learn about the effectiveness of that decision?" Over time, this practice builds an institutional memory of which decisions produced which outcomes, creating the feedback loop that decision intelligence depends on.
Decision intelligence is a discipline, not a tool. The platforms will continue to mature, and the causal modelling capabilities they offer will become more accessible over time. But the foundation, a structural model that connects decisions to metrics to outcomes, is something you can build today. A well-constructed metric tree, designed with decisions in mind, is the most practical step any organisation can take toward becoming genuinely decision-centric. It does not require a seven-figure software purchase or a team of causal inference specialists. It requires clarity about which decisions matter, how those decisions connect to outcomes, and a structure that makes those connections visible to the people who make them.
Continue reading
How to build a data-driven culture
A framework beyond dashboards
Metrics and behavioural science: why measurement changes behaviour
The psychology behind every metric you track
Strategy execution gap: how metric trees bridge strategy and execution
The gap is not between strategy and execution. It is between strategy and understanding.
Build the causal model your decisions deserve
Decision intelligence starts with a structural model that connects decisions to outcomes. KPI Tree helps you build that model as a metric tree: visible, navigable, and owned by the people who make the decisions that matter.
