How to sunset a metric
Organisations are remarkably good at adding metrics. They are terrible at removing them. Over time, dashboards become graveyards of numbers that nobody acts on, definitions that nobody remembers, and KPIs that served a strategy the company abandoned two years ago. This guide provides a practical process for identifying, retiring, and preventing the accumulation of metrics that no longer earn their place.
9 min read
Why metrics accumulate and nobody removes them
Every metric in your system was added for a reason. Someone, at some point, decided this number mattered enough to track. The problem is that the conditions which justified adding that metric rarely persist indefinitely. Strategies shift. Products evolve. Markets change. Teams reorganise. But the metrics remain, because removing a metric requires someone to actively decide it no longer matters, and that decision is far harder to make than the original decision to add it.
There are several forces that drive metric accumulation. The first is loss aversion. People worry that removing a metric means losing visibility into something important, even when they cannot remember the last time anyone looked at it. The second is diffusion of responsibility. Nobody feels personally accountable for the overall health of the measurement system, so nobody takes ownership of pruning it. The third is political inertia. Metrics are often tied to teams, roles, or initiatives. Suggesting that a metric should be retired can feel like suggesting that the work behind it did not matter.
The accumulation problem
The average organisation adds metrics at a far faster rate than it retires them. After three to five years, the typical dashboard contains two to three times more metrics than anyone actively uses. Each unused metric carries a cost: infrastructure to collect it, cognitive load when scanning dashboards, and a subtle erosion of trust when people see numbers they do not understand sitting alongside numbers they rely on.
The result is what some teams call "metric sprawl" or "dashboard bloat." It is not dramatic. Nobody wakes up one morning and realises the measurement system is broken. Instead, the decay is gradual. One more metric gets added during quarterly planning. Another gets inherited from an acquired company. A third was created for a board presentation and never removed afterward. Individually, none of these additions seem harmful. Collectively, they transform a focused measurement system into a noisy collection of numbers that obscures signal rather than revealing it.
Six signs a metric should be retired
Not every metric that has been tracked for a long time needs to be retired. Some metrics remain valuable for years because the business dynamics they measure have not changed. The question is not "how old is this metric?" but "does this metric still earn its place in the system?" The following signals indicate that a metric has outlived its usefulness.
1. Nobody can explain why it is tracked
Ask three people why a particular metric exists, and if none of them can articulate the decision it was meant to inform or the behaviour it was meant to drive, the metric has lost its purpose. This is the single strongest signal that a metric should be retired. A metric that exists without a clear rationale is consuming resources and attention without delivering value.
2. It has never triggered a decision or action
The purpose of a metric is to inform decisions. If a metric has been tracked for six months or more and has never prompted anyone to change course, investigate a problem, or validate an assumption, it is not serving its function. This does not mean the metric was always bad. It may have been useful once and has since been superseded by a better measure.
3. The definition has drifted without anyone noticing
When a data pipeline changes, a product feature is renamed, or a business process is restructured, metric definitions can silently shift. If the number on the dashboard no longer measures what people think it measures, it is worse than useless. It is actively misleading. Definitional drift is especially dangerous because it is invisible to anyone who does not audit the underlying data.
4. The strategy it supported no longer exists
Metrics should be tied to strategic objectives. When the organisation shifts its strategy, the metrics that supported the old strategy should be reviewed. A metric tracking "new market penetration" makes no sense if the company has abandoned its expansion plans. Yet these orphaned metrics persist because nobody connects the strategic change to a measurement change.
5. It has no owner
A metric without an owner is a metric without accountability. Nobody is responsible for checking whether it is accurate, investigating anomalies, or taking action when it moves. Unowned metrics are the most likely to drift, mislead, and accumulate, because there is no one whose job it is to ask whether the metric still belongs.
6. It duplicates another metric with a different name
Over time, especially in larger organisations, different teams create metrics that measure substantially the same thing using different names, definitions, or data sources. "Customer attrition rate," "logo churn," and "account losses" might all capture the same underlying dynamic with minor variations. Keeping all three creates confusion about which is the source of truth.
A useful rule of thumb: if a metric disappeared from every dashboard tomorrow and nobody noticed for a month, it should be retired. The fact that nobody would miss it is the clearest possible evidence that it is not informing decisions.
A step-by-step process for sunsetting a metric
Removing a metric should not be done impulsively. Even metrics that appear useless may have hidden consumers or downstream dependencies. A structured sunsetting process ensures that nothing is lost accidentally and that stakeholders have the opportunity to make a case for keeping a metric before it is retired.
- 1
Identify the candidate
Flag the metric as a retirement candidate based on the signals described above. Document why you believe it should be retired: what purpose it originally served, why that purpose no longer applies, and whether any alternative metric already captures the same information. This documentation is important because it creates a record that future teams can reference if the question arises again.
- 2
Find all consumers
Before removing anything, audit where the metric is used. Check dashboards, automated reports, alert rules, OKR documents, board packs, and any integrations that reference it. You may discover that a metric you thought was unused is actually embedded in a quarterly report that a finance team relies on. This step prevents the most common failure mode of metric retirement: breaking something downstream that nobody anticipated.
- 3
Notify stakeholders and set a grace period
Communicate the intent to retire the metric to everyone who currently consumes it. Give them a clear timeline, typically two to four weeks, during which they can raise objections or identify dependencies you missed. This is not a formality. Stakeholders sometimes have legitimate reasons for keeping a metric that are not visible to the person proposing retirement. The grace period ensures those reasons surface before the metric is gone.
- 4
Archive, do not delete
When the grace period ends without objection, remove the metric from active dashboards and reports but preserve the historical data. Archiving rather than deleting means that if someone later needs to analyse a historical trend or investigate a pattern that predates the retirement, the data is still accessible. The cost of storing historical data is almost always negligible compared to the cost of losing it permanently.
- 5
Document the retirement
Record what was retired, when, why, and who approved it. Maintain a simple log or register of retired metrics. This register serves two purposes: it prevents future teams from re-creating the same metric without knowing it was previously retired, and it provides an audit trail for governance and compliance purposes.
“The goal of sunsetting is not to have fewer metrics. It is to ensure that every metric you keep is one that earns its place by informing decisions, driving behaviour, or revealing problems that would otherwise go unnoticed.”
How metric trees make sunsetting easier
The hardest part of sunsetting a metric is answering the question "does this metric still matter?" In a flat dashboard, every metric has the same visual weight and there is no structural mechanism for distinguishing essential metrics from legacy ones. A metric tree changes this fundamentally by introducing a simple test: if a metric does not connect to the tree, it does not belong.
In a metric tree, every metric exists because it either drives or is driven by another metric. Revenue is driven by new business, expansion, and retention. Retention is driven by onboarding success, product adoption, and support quality. Each metric earns its position by demonstrating a causal or mathematical relationship with its parent and its children. A metric that cannot demonstrate such a relationship has no place in the tree, and that absence of connection is the clearest possible signal that it should be retired.
Consider this tree. Suppose an organisation is also tracking "total registered users," "raw page views," and "employee headcount growth." None of these metrics connect to the MRR tree through a causal relationship. They may be useful for other purposes, but they are not part of the system that explains how revenue works. The tree makes this visible instantly. On a flat dashboard, these metrics sit alongside MRR and its drivers with equal prominence, creating the illusion that they are equally important.
The tree also makes sunsetting decisions less political. When someone asks "why are we removing this metric?", the answer is structural rather than personal: "because it does not connect to any metric in our tree." This shifts the conversation from "do you think this metric matters?" (which invites opinion and defensiveness) to "can you show how this metric drives or is driven by another metric in our model?" (which invites evidence and analysis).
The connection test
Before adding any new metric to your system, ask: "Where does this connect in our metric tree?" If the answer is "it does not," then either the tree is incomplete and needs a new branch, or the metric does not belong. This single question, applied consistently, prevents metric accumulation at the source.
Overcoming political resistance to retirement
Even with a clear process and a structural framework, retiring metrics can meet resistance. Metrics are not just numbers. They are tied to teams, budgets, influence, and identity. When someone proposes retiring a metric, the people associated with that metric may hear "your work does not matter." Navigating this resistance requires both empathy and technique.
| Source of resistance | What it sounds like | How to address it |
|---|---|---|
| Identity attachment | "We have always tracked this. It defines what our team does." | Separate the metric from the team's value. Emphasise that the team's contribution is not diminished by the retirement of a metric that no longer reflects current strategy. Offer to help define a replacement metric that better captures their current impact. |
| Fear of lost visibility | "If we stop tracking this, how will anyone know what we are doing?" | Acknowledge the concern and propose alternatives. The goal is not to make the team invisible but to ensure the metrics they are measured by actually reflect their contribution. A better metric is more visibility, not less. |
| Sunk cost fallacy | "We invested so much in building this metric. We cannot just throw it away." | Remind stakeholders that the cost of building the metric is already spent regardless of whether you continue tracking it. The question is whether it is worth the ongoing cost of maintenance, attention, and dashboard space. |
| Governance concerns | "Compliance requires us to track this." | Take this seriously. Some metrics genuinely are required by regulators, contracts, or audit processes. Verify the requirement independently rather than accepting it at face value, as people sometimes assume a metric is required when it is merely habitual. |
| Uncertainty about the future | "What if we need this metric later?" | This is the strongest argument against retirement and the easiest to address. Archive the data rather than deleting it. Make clear that the historical record will be preserved and accessible. What you are removing is the active tracking and reporting, not the data itself. |
The most effective way to reduce political resistance is to make metric retirement a normal, recurring part of how the organisation manages its measurement system. When sunsetting is treated as an exceptional event, it feels threatening. When it is treated as routine maintenance, no different from updating a strategy document or revising a budget, it becomes unremarkable. The organisations that are best at metric hygiene are the ones that have normalised the question "should we still be tracking this?" as a regular part of their review cadence.
Metric hygiene as an ongoing practice
Sunsetting individual metrics is necessary but insufficient. Without a systematic practice of metric hygiene, the same accumulation problem will recur. Metric hygiene is the discipline of regularly reviewing, pruning, and maintaining your measurement system so that it remains focused, accurate, and aligned with how the business actually operates.
- 1
Conduct a quarterly metric review
Once per quarter, review every metric in your active system. For each metric, ask: does it still connect to the tree? Does it still have an owner? Has anyone acted on it in the last 90 days? Has its definition drifted? This review does not need to be lengthy. An hour per quarter is usually sufficient. The discipline of doing it regularly matters far more than the depth of any individual review.
- 2
Apply the "last looked at" test
If your analytics platform tracks usage, check when each dashboard or metric was last viewed. Metrics that have not been viewed in 90 days are strong candidates for retirement. This data-driven approach removes the need for subjective judgement and provides an objective basis for the conversation.
- 3
Require a business case for new metrics
The most effective way to prevent accumulation is to raise the bar for adding metrics in the first place. Before a new metric is added to the system, require a brief justification: what decision will this metric inform? Where does it connect in the tree? Who will own it? How will we know if it is no longer needed? This does not need to be bureaucratic. A few sentences is sufficient. The act of writing the justification forces clarity about whether the metric is genuinely needed.
- 4
Set expiry dates on experimental metrics
When a metric is added for a specific initiative, experiment, or time-bound project, give it an explicit expiry date. If nobody actively renews it by that date, it is automatically flagged for retirement. This single practice prevents the most common source of metric accumulation: temporary metrics that become permanent by default.
- 5
Maintain a metric register
Keep a central register of every active metric, including its definition, owner, date added, purpose, and position in the tree. This register is the single source of truth about what is being tracked and why. It makes auditing straightforward and ensures that no metric exists without documentation. When a metric is retired, it moves from the active register to the archive.
“Metric hygiene is not about having fewer metrics. It is about ensuring that every metric you track is one you would choose to add today if you were starting from scratch. If you would not add it today, you should not keep it tomorrow.”
The discipline of letting go
The reluctance to retire metrics reflects a deeper organisational tendency: the belief that more information is always better. In reality, more information is only better when it is structured, relevant, and acted upon. Unstructured information is noise. Irrelevant information is distraction. Information that is never acted upon is waste. A measurement system cluttered with retired-but-still-tracked metrics suffers from all three.
The organisations that measure most effectively are not the ones that track the most. They are the ones that track the right things and have the discipline to stop tracking everything else. This discipline is harder than it sounds, because it requires ongoing judgement about what matters, willingness to let go of familiar numbers, and a structure that makes the right answer visible.
A metric tree provides that structure. By requiring every metric to demonstrate its connection to the broader system, it makes sunsetting decisions clear and defensible. By assigning ownership to every node, it ensures someone is accountable for asking whether each metric still belongs. And by modelling the business as an interconnected system rather than a flat list of numbers, it transforms the question from "should we stop tracking this?" into "does this connect to anything that matters?" The answer to the second question is almost always obvious, even when the answer to the first is clouded by politics, inertia, and habit.
The best time to sunset a metric was the moment it stopped earning its place. The second best time is today. Build the practice into your quarterly cadence, embed the connection test into your tree, and treat metric retirement not as a loss but as a sign that your measurement system is alive, evolving, and focused on what actually matters.
Continue reading
Common metric anti-patterns and how to fix them
A field guide to the mistakes that quietly sabotage your measurement system
Vanity metrics vs actionable metrics
The numbers that feel good versus the numbers that do good
Metric ownership
The most underrated lever in business performance
Keep your metrics focused with a metric tree
A metric tree makes sunsetting simple: if a metric does not connect to the tree, it does not belong. Build your tree, assign owners, and let the structure tell you which metrics earn their place and which ones are ready to retire.
