Introduction
A measure is not a vanity mirror, nor is it a scorecard to justify someone’s headcount. It is diagnostic instrumentation. It's only function is to expose friction and systemic failure with complete, unemotional clarity so that people can make hard trade-offs before things break. The moment a measure or target exists to prove that leadership is doing a great job or to protect a comfortable middle manager from accountability, it ceases to be a measurement. It is just an expensive corporate cover story.
Useful measures are diagnostic instruments, not objectives
A measure is not an objective in itself. An objective is the destination, a goal with a clearly defined end state and a deadline. A measure is simply the instrument that measures how far along you are toward the objective, and a way to measure whether you have actually achieved that objective or not.
It does not set strategy, it does not motivate teams, and it should never be treated as an end in itself. Its job is purely diagnostic, it should reflect the ground truth of the system as it runs, without narrative spin or managerial interpretation.
A useful measure is never an academic exercise in statistical gymnastics. If you need advanced math degrees or an explanation of polynomial regression to understand what a number means and how daily work impacts it, the measure is useless.
The best measures are brutally simple, transparent, and immediately obvious to everyone looking at them. If you cannot explain the measure and what drives it in one plain-English sentence, throw it out.
Measures of velocity - Quantitative output
Measure output without apologizing
Management literature has made a habit of sneering at output. We are told by consultants and framework peddlers that "measuring output is an outdated industrial mindset" and that "mature organizations only measure outcomes."
This is at it's best complete nonsense, and at it's worst intellectually dishonest. It is a shield that excuses them for not delivering easily measurable and tamper proof value on a deadline.
Definition of measures of velocity:
- Efficiency gains, such as hours saved on a specific process per month
- Cost per unit produced
- Increased production capacity
These have recently been viewed as less valuable than outcome based ones. This is wrong, a business thrives on the quantitative measures. Focusing solely on qualitative outcomes rather than output, turns an organization into a complicated academic exercise rather than a practical and pragmatic system of systems. Output is the baseline of operational vitality. It is what actually creates value and is the prime mover behind everything else in any organization.
Do not apologize for measuring raw throughput. The goal is not to eliminate output measures, but to ensure that the output you measure is real work rather than administrative motion.
Measures of momentum - Qualitative outcomes
The opposite failure is even more common. An organization can run at full velocity and go nowhere: lines running at capacity on products nobody wants, teams delivering on schedule into a market that has already moved. Output without direction is just activity without effect.
Momentum is velocity with mass behind it, the output that lands, compounds, and becomes true outcomes.
Definition of measures of momentum:
- Quality as the customer experiences it, such as defects reaching customers
- Customer retention and repeat orders
- Revenue and margin from what you produce
- NPS
Measures of integrity
Measures of integrity track the condition of the system itself. They are rarely improved for their own sake. They are held, and most foundational targets are set on them. They can be viewed as foundational measures, things that just has to be working. If they are yellow, the organization has a flu, if they are red, it should be an internal crisis.
- Uptime and outages
- Machines working
- Truck health
- Sick leave and turnover among skilled staff
- Maintenance backlog
- Technical, managerial, and process debt
- Margin distribution across the value chain, and dependence on single suppliers or customers
- technical SEO, such as Lighthouse scores, crawl errors and page speed
None of these will ever make it onto a quarterly highlight slide. That is exactly the point. Integrity is invisible when its green and catastrophic when its red, which is why it must be measured when nothing is wrong, to give an indication before something goes bad.
Paired and counter-weight measures
Pairs of measures should be used to balance the effect and counter-effect, that way we don't overreact to increases.
For instance releases increasing will also increase bugs in production. The bug count needs to be paired with severity, length of bug, economic effect and release frequency.
Every measure you push will push back somewhere else. Pair each primary measure with a counter-weight, so a gain in one place cannot quietly be paid for in another. A measure of velocity should almost always be paired with a measure of momentum or integrity: speed is only progress if it lands, and only sustainable if the system survives it.
Useful targets
A measurement tells you where you are, the target tells you where you need to be. It is a threshold you need to hold or reach. We split it into two types of targets, the time bound and the foundational.
Time-bound targets
Either connected to objectives, initiatives or milestones. These are bound by time and have a clear end-date where it should be measured. It should both be measured during the initiatives and after to see whether we hit the effects we wanted or not. The time bound targets can be ranges (from x to y) as well, or stepped ones (best, worst, realistic).
Foundational targets
These can be bottom line results or top line contributions, or % of FTEs on sick leave at any time, or uptime of production systems. These targets are constant and tied to either business models or the mission itself.
Indicators
Prognoses about whether we are directionally on course or if we are not going to be able to hit the targets that we've set.
Avoid proxy-measures
Proxy-measures sound like good ideas when pitched, but given enough time, they will become gamed, obfuscated and they will end up managing the score instead of managing the reality.
Proxies are not harmless bureaucratic quirks. They are an active drain on organizational energy and focus. Every hour spent negotiating story points, calibrating motion, and sitting in sync meetings to defend synthetic numbers is an hour stolen from fixing real bottlenecks in the actual business or focusing on grabbing the real opportunities. The entire thing becomes a closed loop of fake productivity. A team can hit 100% of its internal proxy targets, win praise across the management chain, and celebrate another "successful" cycle while the core business output quietly degrades.
Instead of defining proxy measures, every measure needs to be put into the same global hierarchy.
Proxies will actively rot the system from within, by measuring movement instead of progress:
- They incentivize systematic gaming by chopping medium things into tiny things, to show ticket velocity instead of feature releases.
- They reward motion over completion: A project coordinator can report an initiative as "85% complete" for three quarters in a row based on checklist proxies, even if zero functioning capability has reached reality.
- They sever teams from reality: When the primary target is maintaining an internal score, nobody has an incentive to ask whether the system is actually holding up under real-world pressure.
Measures and targets need to be tamper proof
Middle managers and coordinators are rational actors. When their status, authority, or quarterly standing depends on a dashboard, human nature takes over and they will bend reality before they admit systemic failures.
Unless you enforce strict, objective governance over how numbers are gathered, measure tampering will run rampant across the organization:
- Redefining terminology on the fly: When incident numbers spike, middle management simply reclassifies critical operational failures as problems instead of incidents to protect the uptime report.
- Filtering the dataset: Negative customer feedback, botched deliveries, or failed transactions are dismissed as statistical outliers and quietly excluded from the monthly review.
- Grading their own homework: Coordinators determine the criteria for whether their own project hit its milestones, ensuring a 100% success rate on paper while the operational floor deals with the wreckage.
Measuring should never be done by the people being measured. If a measure requires human curation, manual entry, or managerial interpretation before it reaches leadership, it has to be done by someone else. Useful measures must be completely tamper-proof. They must be generated automatically by the systems doing the work: raw event logs, verified transaction drops, system latency, and uncurated frontline complaints. If they are qualitative in nature, there should be a separate team doing the qualitative work.
If a middle manager has the administrative access to delete bad feedback or tweak the formula of a scorecard, they will to some degree and at some point in time do it. Protect the numbers from human interference, or stop pretending you care about the truth or are data driven. Then it's better to just have power points with opinions rather than dashboards with fake truths.
Measures cannot be objectives
When a measure becomes a target, it ceases to be a good measure
The moment a target is tied to compensation, performance evaluations, or quarterly status reports, people stop managing the business and start managing the score instead. Measures should tell leaders and managers how far along initiatives are while they run, and how well they solved their stated problems after they are done.