"It is not what you measure... it is what you do with the measurement." Some companies love dashboards. They are televisions scattered across office walls and monthly presentations full of colorful charts, rising numbers, and green arrows everywhere. Looking from afar, it seems like the portrait of absolute success. But when you sit in the meeting and ask a simple question, "so, what changed in our operation after seeing these numbers?", the answer is almost always an embarrassing silence. Metrics exist to guide decisions and correct courses. If numbers go up or down and the team keeps doing the exact same thing on Monday morning, those charts are not management: they are just expensive decoration. ## Vanity metrics vs. action metrics The first trap of any leadership is to confuse raw movement with real progress. It is easy to fall into the temptation of putting only accumulative indicators on the slide that, by mathematical nature, never fall. Classic examples of vanity metrics: - Total number of registered users since foundation: it looks grandiose on the big screen, but how many of these users remain active and generating revenue this month? - Page views or raw clicks: they generated traffic, but did they convert into results or just consume infrastructure for nothing? - Quantity of features launched in the quarter: the team delivered twenty new screens, but did any customer actually adopt these functions or did we just increase system maintenance complexity? Concrete examples of action metrics: - Real conversion rate and retention by cohort: it immediately shows whether the product is delivering the promised value. - Mean time to restore incidents (MTTR): it reveals engineering's true resilience when something fails in production. - Cost per transaction or per active client: it connects the infrastructure bill directly to the business financial margin, as advocated by the FinOps Foundation Unit Economics discipline. - Active transactional NPS: measured right after a delivery or actual service, and not that generic survey triggered once a year just for show. What is the structural difference between the two groups? Action metrics have consequences. If the indicator moves unexpectedly, it triggers a clear investigation and a concrete change in priority. ## Goodhart's Law and Andy Grove's antidote Even when we choose legitimate operational metrics, there is a second danger that every technical leader must master. In 1975, British economist Charles Goodhart formulated in the article Problems of Monetary Management a principle that became world-known as Goodhart's Law (synthesized years later by anthropologist Marilyn Strathern): "When a measure becomes a target, it ceases to be a good measure." In software engineering and operations, Goodhart's Law appears all the time: - If you evaluate a support team only by the average ticket closing time, agents start closing tickets prematurely without solving the customer pain. - If you evaluate developers by the number of commits, story points, or lines of code, the team inflates estimates and breaks simple tasks into ten artificial pieces. - If you demand blind cuts to the gross cloud bill without looking at the operation, someone turns off critical redundancies and drops system availability. How to prevent an important metric from being distorted? In the classic book High Output Management (1983), Andy Grove, legendary former CEO of Intel, taught the most effective technique in management history: paired indicators. Grove shows that we should never measure quantity or speed in isolation; every efficiency metric must be accompanied by a quality counterbalance metric. If you measure delivery speed (lead time), you must obligatorily pair it with the change failure rate. If you measure the unit cost per transaction, you must pair it with latency and the user experience SLA. Thus, no one can "improve" one chart by destroying the other behind the scenes. ## The 4-question test to clean your dashboard If you want to know whether your area's indicators are living management tools or just executive presentation ornaments, run each chart of your dashboard through this four-question filter: 1. If this number goes up 15% this week, what exactly do I do? 2. If this number drops 15% this week, what concrete action do we take? 3. Who is the nominal owner who looks at this number regularly? 4. Was any real product, architecture, or budget decision made in the last 90 days because of this indicator? If the answers to these questions are vague, theoretical, or nonexistent, have the courage to clear the screen: this metric is pure vanity. ## Less is more: 3 to 5 vital signs With modern observability and BI tools, it is tempting to want to measure fifty variables at the same time. But an excess of indicators does not generate control; it generates paralysis and noise. When thirty numbers blink on the screen, none of them is a priority. Focus on the 3 to 5 metrics that truly move the needle in your current context. Choose vital signs that anyone on the team understands without needing a statistical dictionary. When these indicators reach the desired stability level, adjust the scale and evolve the focus. ## The ritual makes all the difference Finally, remember that a metric without a ritual is dead data. A perfect dashboard that is only opened on the eve of the board meeting changes no one's culture. Establish short, regular, and disciplined moments to look at the numbers together with the team. Discuss variations with intellectual honesty, without witch hunts and without trying to makeup the chart when the week is bad. Formulate clear hypotheses, make practical decisions, and, the following week, check in the same data whether the decision made an impact. Measuring and putting it on a slide is easy. Acting with consistency based on what the measurement reveals is the hard part. And it is the only one that truly builds results. > Continue deepening: also read FinOps in 2030: Zero Cost Is Not The Goal and Complete FinOps Guide for Companies. Also test the Tools and Executive ROI Calculator and get to know the practice of FinOps and Cloud Efficiency. --- ### References and verifiable readings - Goodhart, C. A. E. (1975). Problems of Monetary Management: The U.K. Experience. Papers in Monetary Economics, Reserve Bank of Australia / Strathern, M. (1997). ‘Improving ratings’: audit in the British University system. European Review, 5(3), 305-321. - Grove, A. S. (1983). High Output Management. Random House. - FinOps Foundation. Cloud Unit Economics Framework & Capabilities. Linux Foundation.
Good ideas keep the conversation going.
Share this article or publish the visual deck to your Stories and Feed.

Cesar Zanis
15+ years leading software architecture, database performance, FinOps, and applied AI across enterprise operations. Built and scaled an engineering organization from zero to 20+ developers.
Enjoyed this article?
I publish practical analyses on critical systems architecture, Cloud FinOps, data, AI, and engineering leadership directly on LinkedIn.
Continue Reading
Take this analysis further with data and sources
All visual cards of this article are ready above to share on your Stories, Feed, WhatsApp, or LinkedIn with full sources.



