The moment a metric moves is not necessarily the moment the business changed.
The dashboard may show the first visible evidence of a problem today, but the conditions creating that problem could have been in place for weeks or months. A customer was already waiting longer for value. A decision had already been deferred. A handoff had already become ambiguous. Capacity had already been absorbed by work that did not create the intended result.
By the time the metric confirms the issue, the organization may have adapted around it. People create workarounds. Managers add meetings. Teams protect their own queue. Customers change their behavior. The final number is a lagging signal from a system that has already been operating differently for some time.
That is why a response built only around the visible metric can miss the condition underneath it.
The number is late, not useless
Lagging indicators are valuable. They tell you whether the outcome you care about is actually changing. The mistake is asking them to do a job they cannot do alone: explain the sequence that produced the result.
Suppose implementation time increases from eleven months to fourteen. The organization may focus on the accounts that breached the target. That is sensible, but it is not the whole investigation. The more useful questions are when the work began waiting, when scope changed, when a dependency became unclear, when a decision moved from the project team to an executive queue, and when rework started to accumulate.
The metric gives you the doorway. The investigation follows the path to the doorway.
The MIT work on leading indicators describes leading measures as signals that can help predict future system performance before the final result is realized. In a business, a leading indicator might be an unresolved dependency, the number of decisions waiting for an owner, time between handoffs, adoption of a critical workflow, or the proportion of work entering through an exception path.
None of these measures is inherently predictive. They become useful only when the organization can connect them to the outcome it is trying to improve.
Reconstruct the sequence
When an outcome changes, create a simple timeline before creating a new solution.
Start with the first period in which the metric moved. Then work backward through the operating conditions that could have influenced it. Look for changes in:
- customer or product mix;
- staffing, capacity, or leadership;
- pricing, scope, or commercial commitments;
- systems, process, or tooling;
- priorities, incentives, or decision rights;
- dependencies between teams;
- the definition or source of the metric itself.
This is not a request to document every event in the company. It is a way to distinguish an isolated fluctuation from a new operating pattern.
For example, if retention fell in the second quarter, the relevant change may not be a customer complaint in June. It may be an implementation model introduced in January, a product decision that changed the customer’s workflow in February, or a staffing change that moved executive sponsorship from a dedicated owner to an overloaded manager in March.
The closer the team gets to the first condition that made the outcome more likely, the more useful the intervention can become.
Ask leading questions, not only leading metrics
Organizations sometimes respond to a lagging metric by adding a list of new metrics. That can create the same problem at a smaller scale: more observation, no shared interpretation.
A leading question is often more useful than a leading indicator by itself:
- What would have to be true before this outcome improved?
- Which condition tends to move before the result changes?
- Where does work wait before anyone calls it a problem?
- Which decision is repeatedly made too late?
- What is the earliest point where the customer or team can tell that the system is becoming harder to use?
The answers help define what should be measured. If the organization believes margin is worsening because implementation rework is increasing, the leading signal might be the percentage of projects requiring a second scope review. If the belief is that retention is at risk because customers are not reaching value, the signal might be time to first meaningful use rather than a generic engagement score.
The measure should make the driver more visible, not simply make the dashboard busier.
Do not confuse correlation with control
A condition can be associated with an outcome without being a lever.
A team may notice that large customers have slower implementations. That observation may be true, but size alone is not necessarily something the organization can change. The controllable lever may be the scope boundary, the implementation model, the decision process, or the way complex work is staffed.
Likewise, a company may observe that churn is higher among customers who use a particular feature less often. That does not prove that feature usage is the root cause. Customers may use it less because their implementation was incomplete, because the feature does not fit their workflow, or because the relationship is already weakening. The lever could be onboarding, product design, customer fit, or executive alignment.
The goal is not to find the variable most closely related to the outcome. It is to find the condition that is both material and changeable.
Make ownership visible
The time between a system change and a metric change creates a common ownership problem. The people who made the original decision may no longer be in the conversation. The team reporting the outcome may be asked to correct something it cannot control. Executives receive competing explanations because each function is looking at a different part of the sequence.
A good review names the owner of the outcome and the owners of the important drivers. It also makes the handoffs visible. Who can change the condition? Who needs to be involved? What decision is required? How will the organization know that the lever moved before the lagging metric catches up?
This is one reason a short, cross-functional working session can be more valuable than a larger reporting process. The objective is not to gather every stakeholder. It is to bring together the people who can see the sequence and move the lever.
Create a response that matches the lag
Not every metric change deserves a transformation program. Some need a short-term containment action while the deeper driver is investigated. Others reveal a structural issue that requires a redesign.
The response should match the confidence of the explanation. If the driver is uncertain, choose an intervention that generates learning without creating irreversible cost. If the driver is clear and the lever is within reach, define the change, the expected effect, and the review point. If the metric is volatile, set a longer observation window instead of reacting to every movement.
The Harvard Business Review discussion of leading sales metrics makes the broader point that leaders need indicators that help them understand the future value of the business, not only report what has already happened. The same principle applies beyond sales: the useful measure is the one that helps the organization decide earlier and better.
Use the metric as a doorway
The metric matters. It creates the shared fact that a conversation needs. But it is not the whole conversation.
When a number moves, ask what was already true before the movement became visible. Trace the conditions, decisions, handoffs, and capacity constraints that shaped the result. Separate the drivers the organization can influence from the facts it can only observe. Then choose a lever and define the effect you expect to see.
The number tells you where to look. The work is understanding what happened before it arrived.