Stop Describing and Start Prescribing

Written by Ed Cook and Roxanne Brown

Open any change management status report, and you will likely find the same thing: a chart of training completion rates, a table of survey scores, maybe an arrow pointing up or down next to last month's numbers. The presenter will walk through the data, perhaps note that satisfaction scores dipped in the Northeast region, and then ask the room, "Any questions?" There is a pause. Someone asks a clarifying question about the sample size. The meeting ends. No decision is made.

This scene plays out in conference rooms and video calls with remarkable regularity. It reveals a fundamental problem. The analytics have described what happened but have prescribed nothing about what to do next. Pasting an Excel spreadsheet onto a PowerPoint slide is not analysis. It is data delivery. And as we have argued throughout this series on Data-Driven Change Management, data is not information. Information is defined by one single trait: you can make a decision from it.

To move from data delivery to genuine decision support, Change Leaders must climb an analytical ladder. Most organizations are stuck on the bottom rung. The few that climb higher find something unexpected at the top. Not just better analytics, but better insight. The climb alters how Change Leaders think, and that matters as much as the numbers themselves.

The Three Levels of Analytics

Analytics (and therefore Data-Driven Change Management) operates at three distinct levels. Each builds on the one below. Stopping at the first, as most organizations do, is like reading the first chapter of a book on aviation and declaring you know how to be a pilot. The chapter is a complete unit, but it is not everything there is to know.

Descriptive Analytics: What happened?

This is where most change measurement lives and, too often, stops. Descriptive analytics uses probability, basic statistics, and visualization to answer "what happened?" How many people completed the training? What were the survey scores? What is the current attrition rate? This is looking in the rearview mirror of your change. They tell you where you have been. They are useful. You cannot understand where you are without measuring how you got there, but descriptive analytics will not tell you where you are going, and it certainly will not tell you what to do when you arrive. A Change Leader who presents only descriptive data to a senior executive and asks for a decision is placing the entire analytical burden on the executive. That is not decision support. It is abdication.

In our three-tiered metrics approach, Self-Reported, Observable, and Existing Company Metrics, descriptive analytics is the foundation. You need to know, for example, that 72% of respondents said they understand the new process and that help desk calls related to the old system dropped by 15%. Those are facts worth having. They become dangerous only when a leader mistakes them for the complete picture.

Predictive Analytics: What will happen?

The next rung on the ladder shifts from looking backward to looking forward. Predictive analytics uses techniques like regression analysis and simulation to forecast what is likely to happen next. This is the analytical equivalent of switching from the rearview mirror to looking through the windshield.

Consider a practical example. You have collected survey data from pilots in a training program across several phases of your change. Using a Large Language Model (LLM), you can upload that data and ask a plain-language question: "Based on the trend in pilot survey scores over the last three collection periods, what is the projected score at go-live?" The LLM will run a regression, create a predictive equation, and provide a forecast. As we explored in If You Can Conceive It, You Can Calculate It, the ability to ask these questions without writing code has fundamentally lowered the barrier to sophisticated analysis.

There is still an important next level. Predictive analytics tells you where you are heading. It does not tell you what to do when you get there. A weather forecast that predicts rain is useful, but it does not hand you an umbrella. Many organizations that reach this second level, predictive analytics, feel analytically advanced, and relative to most, they are. Yet prediction without prescription still leaves the decision to a mix of intuition, experience, or organizational politics. Those forces are not always wrong. But they are not analytically grounded, and they are subject to what Nassim Nicholas Taleb calls the Narrative Fallacy, the deeply human tendency to construct a compelling story from incomplete evidence because it sounds right. It may be wrong, and it is hard to tell the difference once the story takes hold.

Prescriptive Analytics: What should we do?

The top of the analytical ladder is where analytics applied to Change Management and Decision-Making intersect. Prescriptive analytics does not merely forecast the future. It evaluates alternatives and recommends a course of action. The tools at this level include optimization, game theory, and Decision Analytics. This is the domain where the question shifts from "what will happen?" to "what is the best thing to do?"

This is fundamentally a different kind of analysis because the output is not a number or a trend line. The output is a recommendation.

Consider our pilot training example extended to its full potential. Descriptive analytics told you that survey scores dipped in the second phase. Predictive analytics told you that, if the trend continues, readiness at go-live will fall below your safe threshold. Prescriptive analytics evaluates three alternatives: accelerate the coaching program for the lowest-scoring cohort, delay go-live by two weeks to allow additional practice time, or redesign the training sequence based on which modules had the weakest scores. The prescriptive model weighs each alternative against your defined success criteria and tells you which one optimizes across those values.

As with every application of AI in our Data-Driven Change process, the LLM does the calculating. The human leader does the thinking. You must define the values against which alternatives are weighed. AI cannot do that for you. You must provide the context that the model cannot see. Once you do, the analytical engine can evaluate more alternatives, across more variables, more rapidly than any team of analysts working with spreadsheets. If you can conceive of it, you can calculate it. And at the prescriptive level, what you are conceiving of is not just a measurement. It is a decision.

The Burden and Relief of Choice

There is something comfortable about descriptive analytics. A dashboard that shows training completion at 72% does not demand anything of the leader who reads it. It informs. It may even alarm. But it does not force a choice. The leader can nod, express concern, and move to the next agenda item.

Prescriptive analytics removes that comfort. When the analysis says, "Given your stated values and the current data, the optimal path is to delay go-live by two weeks and redirect coaching resources to Cohort 3," the leader is no longer observing. The leader is choosing. Accept the recommendation and own the consequences. Reject it and own a different set of consequences. Either way, a decision has been made visible in a way that descriptive data never requires.

This is where the analytical evolution becomes a leadership evolution. One of our Core Four Philosophies states: 

Before change can happen, the leader must change first.

At the descriptive level, a leader asks, "Are we on track?" At the predictive level, a leader asks, "Where are we heading?" At the prescriptive level, a leader asks, "What is the optimal decision to get us where we need to be?" That last question is harder. It requires the leader to define what "where we need to be" actually means in terms of values, not just targets. It requires the willingness to act on what the analysis reveals, even when the recommended path is not a politically convenient one.

That shift is difficult, but it is also, for many leaders, a relief. The prescriptive model does not eliminate judgment. It gives judgment a foundation. We still need leaders, because the models are not perfect. The aphorism: “All models are wrong, but some are useful,” applies here. Instead of analyzing twenty data points in their head while twelve people watch, the Change Leader has a structured framework that already does the analysis. Better yet, it guides them to the synthesis, where the prescriptive action becomes clear. The decision is still theirs, but it is no longer a guess dressed up in data.

A descriptive analytics dashboard is a speedometer. It tells you how fast you are going. Predictive analytics and prescriptive analytics together are the navigation system that tells you exactly when your trip is predicted to be longer and how to reroute when the bridge is out.

Two Sides of the Same Coin

We named our company "The Change Decision" for a reason. At the analytics conferences, where Ed would go, no one wanted to talk about his fancy math. They wanted to talk about how to “get people to accept and implement” the decision. They wanted to talk about change management. At the change management conferences, where Roxanne would go, people would lament the “bad decision” that they now needed to implement. They wanted to talk about decision-making.  It was as if the groups were in adjacent rooms having parallel conversations, separated by a wall that did not need to be there. The Change Decision got its name because we wanted to remove that wall.

Our work at The Change Decision spans three areas: Change Management, Decision-Making, and Team Effectiveness. The three levels of analytics are the connective tissue between the first two. When a change leader moves from descriptive to prescriptive analytics, they are simultaneously becoming a better decision-maker. The frameworks are not borrowed from one field and applied to the other. They are the same framework viewed from different angles.

The practical implication is this: if your change analytics do not set up a clear decision, they are incomplete. Ask yourself the test we have used throughout this series. Can I make a decision from this analysis? If the answer is no, you have not climbed high enough on the analytical ladder. You are still describing. The data may be accurate. The visualizations may be polished. But the analysis is not finished until it prescribes or at least differentiates between alternatives.

For many Change Practitioners, prescriptive analytics feels out of reach. It sounds like something that requires a data science team and a seven-figure budget. It did once. With modern LLMs, the barrier is the same as it has been throughout this series: willingness. Change Leaders who demonstrate the willingness to ask a clear question, provide clean data, and let the tool do the calculating while they do the thinking can move their teams through change with far more success. The models can now run optimization routines, evaluate competing alternatives, and present trade-off analyses in plain language. The expertise required is not in operating the tool. It is in knowing what question to ask and what values should govern the answer.

That is not a technical skill. It is a leadership skill. And it is a skill that grows when exercised. When Change Leaders and Change Practitioners move beyond the comfortable certainty of "here is what happened" into the productive discomfort of "here is what we should do and why."

Start Prescribing

The Change Practitioner's analytical journey mirrors the change curve itself. In our Change as Experienced Model  (Engage > Understand > Test & Learn > Adopt), most practitioners are somewhere between Engage and Understand when it comes to prescriptive analytics. They know it exists. They may even see its value. But they have not yet had the chance to test and learn with it in a real context.

That opportunity is here. The tools are accessible. The analytical ladder is not a credential gate. It is a practice. Start with the descriptive data you already have, the training scores, the survey results, and the adoption metrics. Ask the LLM a predictive question: "Based on these trends, where will we be at go-live?" Then push further. Ask a prescriptive question: "Given these three alternatives, which one optimizes for both readiness and schedule?" You may be surprised at how quickly the analysis moves from interesting to useful.

Stop settling for dashboards that describe the past. Start demanding analytics that prescribe a path forward. The data your team has collected deserves to do more than fill a status report. It deserves to drive a decision.

The other tools in the toolbox will give you the complete set. The question and hypothesis give you direction. The minimum set of metrics gives you efficiency. The timing of your collection gives you relevance. And prescriptive analytics gives you the thing that all of those were building toward: a clear, defensible, values-based decision. That is the promise of Data-Driven Change Management. That is the guide to finding Joy at Work.