What Sits Upstream of the KPI

Most strategies eventually become numbers. A decision to expand into larger companies becomes an enterprise revenue target. A commitment to customer engagement becomes a retention goal. A plan to build a new category becomes an adoption metric. From there, strategy enters the operating system of the company through Objectives and Key Results (OKRs), Key Performance Indicators (KPIs), dashboards, business reviews and individual goals.

That translation is necessary. Numbers make ambition observable and create accountability. But they also compress the reasoning that produced them. A retention target may begin with a belief about which customers create durable value. A margin target may reflect a choice about which forms of growth are worth pursuing. Once that reasoning disappears, the metric can remain visible even as its meaning fades.

This matters even more as AI makes analysis, forecasts and dashboards abundant. A KPI is not merely a target. It has a strategic hypothesis behind it and a feedback loop around it. The first gives the number meaning. The second allows experience from the field to test whether that meaning still holds.

The Story Behind the Number

Kenneth Boulding offered a useful way to think about the first part. In his 1956 book The Image: Knowledge in Life and Society, the economist and systems thinker argued that people behave according to the picture of reality they carry in their minds. That picture includes experience, expectations, beliefs, relationships and values. Boulding called it “the image”.

His central proposition was simple: “Behaviour depends on the image.” He then added something equally important: “The meaning of a message is the change which it produces in the image.”

This provides a useful lens on strategy. Imagine that a company sets an OKR to increase enterprise retention from 85% to 92%. For the CEO, that number may represent a move towards higher-quality recurring revenue. For product, it may imply deeper usage of a core workflow. For customer success, it may mean improving implementation during the first six months. For sales, it may influence which customers the company pursues in the first place.

The role of narrative is to connect these views into a shared strategic picture. A fuller expression of the same retention goal might read like this:

We believe enterprise customers create the greatest lifetime value when they reach meaningful usage within the first 90 days. Over the next year, we will organise sales qualification, onboarding, product activation and customer success around that moment. We expect activation to improve first, deeper usage to follow, and retention to strengthen over time.

The organisation now has more than a target. It has a causal story. If activation improves while retention remains unchanged, the company has evidence about where it may be in the journey. If retention rises while depth of usage moves differently, that becomes a strategic question worth investigating. The metric sits inside an explanation of how the business believes value is created.

Amazon has institutionalised this relationship. Its written narratives and Working Backwards process require teams to explain the intended customer value before reducing it to milestones and measures. The larger point is that narrative gives data strategic memory. It preserves the beliefs that produced the metric, allowing the organisation to examine both the result and the reasoning behind it.

Narrative gives data strategic memory. It preserves the beliefs that produced the metric.

The danger begins when that reasoning falls away and the KPI starts to substitute for the strategy. Teams optimise the number rather than the outcome it represents. Contradictory evidence gets filtered out because it does not fit the dashboard. Precision creates confidence without necessarily creating understanding.

A dashboard records what moved. It cannot explain why it moved or whether that confirms the original belief. Leaders must hold the metric and its causal story together. They must also look beyond the dashboard, where some of the earliest evidence that a strategy needs to change will appear.

From Cascade to Conversation

A salesperson discovers that customers are using a product for an unexpected purpose. A store manager notices repeated requests for a colour or design that does not exist. A plant worker spots a pattern before it appears in an aggregate quality metric.

These observations begin as local knowledge. Their strategic value grows when the organisation can combine them, interpret them and carry them back towards the people responsible for the larger picture. The question is not simply whether the organisation collects more data. It is whether evidence from the field can alter the strategic response.

Boulding’s image helps here too. The image guides behaviour. Behaviour generates new messages from the world. Those messages modify the image. Strategy, viewed this way, is not a statement that moves once from the centre to the organisation. It is a picture that must keep changing as the organisation encounters reality.

Toyota shows how such a return path can be built into an operating system. Within the Toyota Production System, the principle of jidoka lets operators surface an abnormality immediately. A worker can activate an Andon signal and production can stop so the issue is understood and corrected. The mechanism gives the frontline the authority to interrupt the process rather than let a problem disappear into an aggregate measure.

That authority matters. When people closest to customers or operations know that an unusual observation will receive serious attention, they surface it earlier. When leaders show that frontline information can influence a decision, the organisation builds a richer sensing capability. Trust, in this context, improves the information system itself.

The familiar language of strategy execution is the language of cascading. Strategy cascades into objectives. Objectives cascade into KPIs. KPIs cascade into team and individual accountability. But cascade describes only one direction of movement. Strategy also needs conversation.

The organisation explains the belief behind the metric. Teams act on it. The field generates evidence. Trusted channels carry that evidence upward and across functions. Leadership interprets it against the original narrative, and the next actions reflect what the organisation has learned. The loop runs from belief to measure and action, and then back through field insight to a revised belief.

Cascade describes only one direction of movement. Strategy also needs conversation.

A business review can then examine performance and learning together. Alongside "Did we hit the KPI?", leaders can ask: What did we originally believe would create this outcome? What are the numbers and the field telling us about that belief? What have we learned that should influence the next decision?

What AI Adds to the Loop

This is where AI changes the possibilities. A large company may have thousands of customer conversations, support interactions, sales notes, survey responses, operational exceptions and competitor signals every week. Historically, much of that qualitative information remained local because synthesising it at scale required enormous effort.

AI can bring structure to that field intelligence. It can identify recurring themes across sales calls, connect support tickets to changes in usage, detect new language in customer conversations, compare frontline observations across regions and link those signals to movements in operating metrics. Evidence that was once scattered across the organisation can become visible to the people shaping strategy.

That is a change in governance, not merely an improvement in analytics. Leaders can examine not only whether a KPI moved, but also what customers and employees were experiencing as it moved. The qualitative and the quantitative can begin to inform each other at a scale that was previously difficult to imagine.

But the same capability introduces a new risk. AI can flatten anomalies, overstate patterns and give weak signals a deceptive coherence. It can make the field more audible; it cannot decide which voices deserve attention, whether the original hypothesis still holds or what should change. Those remain questions of judgement.

The first essay in this series argued that when intelligence becomes abundant, leaders need stronger judgement about what matters. The second explored the enduring beliefs that give that judgement a baseline. The mechanism that connects those ideas to the daily life of the organisation is this loop between belief, measure, action, evidence and revised belief.

Strategy needs measurement because choices eventually have to become commitments. It needs narrative because people must understand what those commitments mean. It needs trust because evidence from the field must be allowed to challenge the original assumptions. And it needs judgement because neither a metric nor an AI-generated synthesis can decide what the organisation should continue to believe.

The operating unit of strategic management, then, is not the KPI alone. It is the conversation that sits around it—and the learning that travels back upstream.