We live by numbers. GDP, rankings, dashboards, test scores, algorithms. They feel precise, objective—beyond argument. But what if that certainty is an illusion? In this wide-ranging conversation, Dr Sumit Chowdhury makes a provocative case in Zen and the Practice of Precision, a book he’s working on: measurement is not just technical, it is moral, political, and deeply human.
Dr Chowdhury is former president of Reliance Jio, senior leader at IBM and KPMG, the founder of Gaia Smart Cities, and an advisor to governments and enterprises on digital transformation and responsible data systems.
Drawing on science, history, and AI, he shows how breakthroughs emerge from “beautiful mistakes,” how dangerous metrics have shaped power and policy, and why artificial intelligence is quietly inheriting decades of flawed measurement. The real crisis, he argues, is not data—but how we define, design, and trust the numbers that govern our lives.
Key Takeaways
(Read Time: 4 mins)
1. Numbers are not neutral—they are arguments
“Measurement is a search for truth… Numbers aren’t mirrors, they are choices.”
We treat numbers as facts, but every metric is the result of human choices: what to count, how to count, where to draw boundaries, and when to stop measuring. Measurement is not a mirror of reality; it is a negotiated version of truth. When those choices go unquestioned, numbers quietly accumulate power. We need to ask better questions with our measurements.
History is full of breakthroughs that often start by that awkward reading that you were getting when you were measuring something.
2. Measurement is a moral and political act
“It shapes how we determine something. It's an argument or a compromise between two people to decide what is right, what is wrong”
From calendars and timekeeping to GDP and poverty lines, standards have historically been set by those in power. Metrics allocate dignity, money, attention, and legitimacy. When measurement goes wrong, it doesn’t merely misinform, it reshapes societies, often with irreversible consequences.
3. Anomalies are where new science is born
“Anomalies are like the debug logs from the universe… The fact that you didn’t get data is data.”
Some of humanity’s biggest breakthroughs emerged not from clean data, but from stubborn anomalies: null results, misbehaving curves, contaminated experiments. When measurements refuse to align with theory, they force us to rethink reality itself.
True precision lies as much in observation as in instrumentation. Whether in physics or biology, breakthroughs came from scientists who paused, looked harder, and asked why something felt “off.” Curiosity, not compliance with models, is the engine of discovery.
4. Max Planck and the ultraviolet catastrophe
A lot of today's science came from the fact that we understood light. Early theories treated light as a continuous wave and predicted infinite energy at ultraviolet frequencies—a conclusion the data stubbornly refused to support. The measurements were right; the theory was wrong. Max Planck resolved this mismatch by proposing that energy comes in discrete packets, or quanta. Once the theory changed, the data made sense—and modern physics was born. From this “curve that wouldn’t behave” emerged quantum theory, semiconductors, lasers, MRI, and much of contemporary science.
5. Have the curiosity to ask questions about the measurements one is seeing
Penicillin was born not from precision instruments but from careful attention. When Alexander Fleming returned to a contaminated petri dish, he noticed a pattern others would have discarded: bacteria dying around a patch of mould. That act of observation—not measurement—led to the discovery of penicillin, later scaled during wartime to save millions of lives. The lesson is clear: breakthroughs don’t always come from eliminating noise. Sometimes, curiosity and the willingness to ask what an anomaly is trying to tell us make all the difference.
6. Dangerous metrics turn measurement into theater
“When the measurement is wrong and the incentives are perverse, the metrics become weapons.”
Phrenology, IQ tests, crime statistics, and single-number indicators gained authority by appearing objective. In reality, they were shaped by biased samples, flawed assumptions, and political incentives—and went on to legitimise exclusion, segregation, and control. When metrics stop being questioned, they turn into weapons.
7. What we measure changes behaviour—and reality
“Numbers can create the problems that they claim to measure.”
For example, arrest data measures policing, not crime. When arrests are treated as a proxy for crime, they create self-fulfilling feedback loops: more policing leads to more arrests, which then “prove” that crime is rising. Over time, the metric reshapes behaviour and reinforces itself. You have set a KPI, and you have set more police people down there, they will arrest more. Measurements have feedback loops: the first measurement changes the second measurement, and the second measurement changes the third measurement—the linked measurements create a pattern.
Every measurement also has an uncertainty band—and if the uncertainty is too big, the data is noise.
8. Single “golden numbers” flatten complex truths
“If you define the basket too thinly, the poor disappear completely.”
GDP, poverty lines, AQI, rankings—all reduce multidimensional realities into neat aggregates. In doing so, they hide distributions, uncertainties, unpaid labour, ecological costs, and lived experience. Definitions become policy, and entire populations can disappear on paper.
For example, poverty and hunger are not objective facts but outcomes of definition. Where we draw the line—what counts as poverty, how hunger is measured, and which questions are asked—determines who appears poor and who disappears from the data.
9. AI is inheriting our oldest measurement flaws
“We are training our AI with data from the past that lacks precision information and context.”
Artificial intelligence is being trained on data that is devoid of context, uncertainty, and precision. A “5” is treated as a universal truth, regardless of how, where, or why it was measured. As models themselves become measurement instruments, flawed assumptions get automated and amplified at scale.
10. Measurement design matters
Our models themselves are now meters. They are measuring credit risk, toxicity, attention, stress. These are all models that somebody has created, and they are themselves a meter. Are we going to trust that meter?
Multi-billion dollar decisions are made based on the measurement; we need to very clearly articulate what we need to measure, how we need to measure, the design of the measurement.
Everybody does the definition, but in time, people forget how we came up with that number.
11. Zen, humility, and the courage to revise
“Have the courage to change it when new data comes in.”
It is ultimately about disciplined awareness: the humility to accept that we may be wrong, the mindfulness to revisit assumptions, and the courage to retire harmful metrics. That good governance (of the measurement process) is about knowing when—and how—to change what we measure.
To CEOs and policymakers: look at the consequences of errors made based on that decision, and have the governance process in place to retire harmful metrics.
12. Numbers should invite conversation, not end it
“Numbers are not destiny. They are proposals.”
When metrics become unquestionable—dictator numbers that shut down debate—societies fracture. Numbers should remain proposals, not destinies: clear enough to act on, kind enough to account for consequences, and open enough to be challenged.
